Posted on

Reassessing UK law enforcement data adequacy

The UK government says reforms to police data protection rules will help simplify law enforcement data processing, but critics argue the changes will lower protection to the point where the UK risks losing its European data adequacy.

Currently going through the committee stage of Parliamentary scrutiny, the Data Use and Access Bill (DUAB) will amend the UK’s implementation of the European Union (EU) Law Enforcement Directive (LED), which is transposed into UK law via the Data Protection Act (DPA) 2018 and represented in Part Three of the act specifically.

In combination with the current data handling practices of UK law enforcement bodies, the bill’s proposed amendments to Part Three could present a challenge for UK data adequacy.

The DUAB changes the law to allow routine transfer of data to offshore cloud providers, remove the need for police to log justifications when accessing data, and enable police and intelligence services to share data outside of the LED rules.

In June 2021, the European Commission granted “data adequacy” to the UK following its exit from the EU, allowing the free flow of personal data to and from the bloc to continue, but warned the decision may yet be revoked if future data protection laws diverge significantly from those in Europe.

While the government argues that its reforms will simplify police data processing, critics say the proposals represent enough of a divergence from EU law that it will likely undermine the UK’s LED adequacy.

They add that many of the government’s changes to police data protection rules are a response to a widespread lack of compliance with key provisions in the DPA 2018, such as the need to log justifications when accessing data or implement controls that limit the offshoring of sensitive law enforcement data to non-law enforcement bodies, including cloud providers.

Computer Weekly contacted the Home Office about every concern raised, and the threat to the UK’s LED adequacy created by the government’s proposed changes to the law enforcement data protection regime.

“We have introduced some targeted amendments in the Data Use and Access Bill to improve public trust and to drive up law enforcement efficiency by simplifying the legislation. We are committed to data adequacy and had the UK’s adequacy decisions in mind when producing this bill,” said a spokesperson.

“Any changes to our data protection regime must not come at the expense of security, and high standards of protection will continue to be applied.”

The adequacy process

In exiting the EU, the UK became a “third country” under the bloc’s rules, which means the European Commission (EC) will have to periodically assess whether the country’s data protection framework and practices provide an essentially equivalent level of protection for EU citizens’ data.

The EC will therefore have to make two separate adequacy determinations under both the General Data Protection Regulation (GDPR) and LED by the end of June 2025.

Data protection experts previously claimed to Computer Weekly in February 2021 that any adequacy decision made under the LED would be principally political in nature if it fails to directly address how the data practices of the UK’s criminal justice sector and intelligence services undermine the data and fundamental rights of EU citizens. If this is not addressed, they said a positive adequacy decision could be open to legal challenges in the European courts.

In October 2024, the UK Parliament’s European Affairs Committee (EAC) – in a warning about the risks of the UK losing its data adequacy – highlighted many of the same issues as the experts Computer Weekly spoke to, noting these would be of “interest and potential concern” to both the EC and European Court of Justice (CJEU) as they consider the UK’s adequacy statuses.

This includes potential divergence on data protection standards that would make it harder for people to exercise their data rights; the possibility that the UK government undermines end-to-end encryption; the independence and effectiveness of the Information Commissioner’s Office (ICO); aspects of the UK’s national security regime under the Investigatory Powers Act 2016, including data collection and retention, surveillance powers and practices, and the role of the Investigatory Powers Tribunal; and any legal cases which provide grounds for concern about UK data protection standards.

The EAC also highlighted potential risks posed by onward transfers of data from the UK to other third countries, including under the UK-US Cloud Agreement.

However, the EAC’s findings were published a day before the DUAB was announced, and two days before the text was published online, meaning its inquiry focused on the previous government’s Data Protection and Digital Information (DPDI) Bill – which was dropped from the legislative agenda during the UK’s pre-general election “wash up” period.

While the EC’s adequacy decision will rest on the exact contents of DUAB – for which there is still no official Keeling Schedule – it will be looking to assess whether the framework provides an essentially equivalent level of data protection for EU citizens’ data.

While some of the more controversial measures contained in the previous DPDI Bill – including removing the need for data protection impact assessments and abolishing the dual biometrics and surveillance camera commissioner role – have been dropped in the DUAB, many aspects of it have been carried over.

There are also a number of new measures that may create fresh adequacy-related problems, particularly changes to the international data transfer regime for police.

While an amendment to the DUAB was tabled by Liberal Democrat peer Lord Clement-Jones that would have required the secretary of state to carry out a formal impact assessment of the bill concerning the UK’s data adequacy, government ministers argued against it during the Lords first committee stage on 16 December 2024.

Responding to Clement-Jones during that debate, Baroness Jones, parliamentary under-secretary of state at the Department for Science, Innovation and Technology (DSIT), said maintaining adequacy was a priority for the government, noting that the free flow of personal data with the EU is vital to research, innovation and safety.

“For that reason, the government is doing all that it can to support its swift renewal. I reassure noble Lords that the bill has been designed with EU adequacy in mind,” she said.

“The government has incorporated robust safeguards and changed proposals that did not serve our priorities and were of concern to the EU. It is, though, for the EU to undertake its review of the UK, which we are entering into now. On that basis, I suggest to noble Lords that we should respect that process and provide discretion and not interfere while it is underway.”

A similar position has been adopted by information commissioner John Edwards, who in response to the DUAB said: “Whilst ultimately a decision for others, in my view the proposed changes in the bill strike a positive balance and should not present a risk to the UK’s adequacy status.”

However, the position of the UK government and ICO differs significantly from the views of a number of specialists familiar with both the EU LED and the UK DPA Part Three. Computer Weekly contacted the Home Office about what robust safeguards have been put in place, and which DUAB proposals have been changed that were of concern to the EU, but received no response on this point.

National security or law enforcement?

Chris Pounder – director of data protection training firm Amberhawk – wrote in a blog post that the DUAB would allow the secretary of state to designate that certain police datasets can become subject to Part Four national security rules, rather than Part Three law enforcement rules, over which the ICO has limited enforcement powers.

“The proposal has the effect of taking large volumes of personal data out of the UK’s data protection regime,” he wrote.

Part Four processing is also completely separate from the LED or GDPR and has no equivalent in EU law, effectively lifting police data out of the scope of EU law in instances where the secretary of state decides police and intelligence bodies can share the data.

The [DUAB] proposal has the effect of taking large volumes of personal data out of the UK’s data protection regime Chris Pounder, Amberhawk

Computer Weekly contacted the Home Office about the removal of policing data from the data protection regime, but received no on-the-record response on this point.

Pounder further noted that while the ICO is being abolished in favour of the “Information Commission”, the problem remains in the DUAB that the secretary of state will be able to appoint the most important members of the Commission, which has the potential to give them undue influence over the new body’s decision-making processes.

“The Commission still has to have regard for: the desirability of promoting innovation and competition; the importance of the prevention, investigation, detection and prosecution of criminal offences; and the need to safeguard national security,” he wrote. “In other words, these ‘regards’ could fetter decisions to protect the privacy of data subjects.”

Pounder added the DUAB will also permit the secretary of state to apply a “data protection test” when considering whether a country, part of a country, or a controller located in a country offers an adequate level of protection.

He said the provisions will increase the risk of divergence from EU transfer standards if the EC and UK government have differing views on what “adequate” means here. “Also I don’t understand how a country is not deemed adequate, but a controller, processor, or recipient located in that country is,” Pounder added.

While the UK has already taken steps to award its own law enforcement adequacy to countries not recognised by the EU – including the Isle of Man, Jersey and Guernsey – the EU has not yet reacted to these changes.

Thomas Barrett, a partner at CyXcel who leads the organisation’s data protection and privacy practice, and has previously advised the Home Office and Ministry of Justice on compliance with the DPA 2018, said there are certain scenarios where specialist police units within forces may have to collaborate with intelligence services for particular operations – for example, in terrorism cases where intelligence services have information but no power of arrest as police do – adding while “it raises red flags … I would be surprised how many of these are made”.

He added that in cases where this power is used, it has the potential to be “more targeted, more proportionate, and safer,” because only one set of data protection requirements would apply to this processing, rather than potentially three currently.

As a result, Barrett said the changes being made to UK law via the DUAB are very unlikely to materially affect the country’s LED adequacy.

“It would be counter-productive to remove adequacy over such small changes … there’s so much [law enforcement] cooperation. … Looking at the detail, I struggle to see how you really make hay of a lot of it.”

He said the real risk to LED adequacy therefore lies at “the political level”, which will be decided between the EC and the UK government.

Law enforcement transfers

Independent privacy consultant Owen Sayers, a long-term commentator on DPA Part Three compliance issues with more than 25 years of experience in delivering secure solutions to policing and the wider criminal justice sector, said for the first time UK legislation would place individual data processors – such as cloud providers – on the same broad footing as overseas law enforcement organisations, exempting them from the list of mandatory transfer conditions outlined in Article 39 of the LED.

This includes that the transfers be strictly necessary, that no data subject rights override the public interest of the transfer, that transferring to another policing body – or “competent authority” in LED parlance – would be ineffective, and that the controller provides specific instructions of how to process the data in that particular case.

Under the UK’s current law enforcement-specific data protection rules, police data controllers are bound by the DPA 2018’s stringent transfer requirements, which fully mirror EU law.

This means that, as it stands, each individual law enforcement data controller must ensure that a contract in writing exists between itself and the data processor, which sets out details of the processing, including its duration, nature, and the type and categories of personal data involved. To be valid, the contract or terms of service must be explicit in how they meet the DPA requirements.

Police data controllers are also required to ensure the processor seeks and receives permission before transferring data to a third country, for each particular transfer made. This means each transfer must be assessed on a case-by-case basis.

Police data controllers are further required to perform a case-by-case analysis and justification for all personal data offshored to such processors, and to report this to the ICO. Although police forces have used Microsoft and Amazon Web Services services for the past six years – meaning millions of these transfers will have taken place – the ICO revealed in a Freedom of Information (FoI) response to Sayers that only 148 such notifications had been received up to June 2023.

As previously reported by Computer Weekly, the use of hyperscalers under current UK law presents a number of data protection concerns, including US government access via the country’s invasive surveillance laws, and an inability to comply with the strict transfer requirements contained within the DPA 2018.

In June 2024, Computer Weekly reported details of discussions between Microsoft and Scottish policing bodies – obtained via FoI rules – in which the tech giant admitted it could not guarantee the sovereignty of UK policing data hosted on its hyperscale public cloud infrastructure.

As a result of these FoI responses, Sayers said the law is breached far more often than it is adhered to: “The evidence to show that multiple parts of the Part Three legislation are consistently breached or simply ignored by policing and their justice partners is overwhelming. In truth, the number of organisations who do apply the law as it’s currently written is less than a handful, though those that do so do it very well.”

Mariano delli Santi, legal and policy officer at the Open Rights Group (ORG), said these issues mean it is an open question whether cloud providers can adhere to Part Three requirements in practice. “Given the issues around sovereignty, is a cloud provider able to enforce the contractual agreements entered into with the police? I think that’s an issue that would cause concern,” he said.

Since the re-election of Donald Trump, delli Santi pointed out that the US government has broken several adequacy-related commitments made to the EU around enhancing scrutiny and ensuring the proportionality of their intelligence services operations.

“The Trump Administration fired members of the Privacy and Civil Liberties Oversight Board, and then doubled down with the Federal Trade Commission. Both bodies were fundamental pieces of the EU-US Data Protection Framework [DPF] which, at this point, is quite certain to be struck down by the CJEU,” he said, adding the UK-US Data Bridge, which acts as an extension of the DPF, will also go down if the EU invalidates the framework.

“It has now become obvious that the EU-US DPF will not last for long, and it has just as obviously become unfeasible to rely on US cloud providers for storing personal data unless you are willing to compromise the security and sovereignty of the data you transfer. Indeed, European lawmakers have already started to discuss this.

“Based on all the above, it is now a fact that relying on US cloud services constitutes a threat to the sovereignty, security and autonomy of the UK. Until now, this has been treated as a risk-mitigation issue at best, or something to be swept under the carpet at worst.”

Highlighting the lack of clarity from the UK data regulator around cloud data sovereignty and the applicability of standard contractual clauses in this context, delli Santi said this has created a grey area in which transfers have been allowed to continue.

“The UK government, on their side, have tried to formalise this approach with the DUAB, which introduces a new data transfer regime specifically designed to accommodate the ICO’s ‘tolerant approach’ toward data transfers that lack effective safeguards, and allow data transfers to countries such as the United States by sidestepping human rights and data security concerns.

He added that “the UK needs an exit plan to progressively cut reliance on US digital infrastructure and services – and we need this plan fast”, which includes contingencies to move away holding companies or subsidiaries of US firms geographically based in Europe, which still fall under US jurisdiction.

Given the issues around sovereignty, is a cloud provider able to enforce the contractual agreements entered into with the police? I think that would cause concern Mariano delli Santi, Open Rights Group

“Any of these companies are under an obligation to cooperate with law enforcement and international security authorities in the United States, which can be ordered to hand over data without necessarily having to tell the contracting party,” said delli Santi.

According to the government’s explanatory notes published for the DUAB in October 2024 (paragraph 1022), Schedule 8 of the bill seeks to widen the transfer conditions “by expanding the list of intended recipients to specifically include processors acting on behalf of, and in accordance with a contract with, a controller”.

It added that while transfers to processors in third countries are currently permissible, “this amendment clarifies the existing law and provides legal certainty to UK controllers that they can transfer personal data to their processors operating outside of the UK”.

The explanatory notes also specify that the DUAB will no longer require “controllers to notify the commissioner on each occasion data is transferred; it simply requires notification of the categories of information” that will be transferred.

However, Sayers argued that even if the US government does utilise its various surveillance laws to gain access to UK data, the transfers would be unlawful anyway as UK law lays down a series of specific steps that must be followed for each and every transfer of a specific piece of personal data under Part Three.

“These steps are not being followed, and Microsoft has made clear that they cannot be followed – actually, they’ve said ‘impossible to operationalise’. Because the steps laid down in the DPA 2018 Part Three are not and cannot be followed, that is one of the main reasons why the processing being done on these clouds is in breach of UK law,” he said.

“It makes zero difference if the US government bogeyman tries to use the Cloud Act to look at the data or not, as the data was illegally transferred regardless of the Cloud Act.”

The steps laid down in the DPA 2018 Part Three are not and cannot be followed [which is] one of the main reasons why the processing being done on these clouds is in breach of UK law Owen Sayers, independent privacy consultant

He added: “The intention [of the new DUAB] is to put non-UK processors – principally hyperscalers – on the same broad legal footing as overseas law enforcement organisations.”

He pointed out that the bill would enable UK policing bodies to send data overseas to offshore processors with minimal restrictions. “The bill actually puts overseas processors above overseas law enforcement processors, in the respect that it completely removes obligations to record what data is transferred to them, inform the ICO or make any assessments as to whether a particular transfer is safe and consider the data subject’s rights in advance of sending the data.”

Sayers added that while these and other changes to Part Three would be directly contradictory to EU law, the most likely outcome would be the CJEU finding that the UK regime falls far below EU standards and thus moves to block UK data transfers.

He further added that individual member states may also deem UK laws to be too divergent from their domestic laws to continue to send data, noting the chance of this is high given there are 27 member states, each with their own implementation of the LED.

“You can 100% use cloud for law enforcement data, but it needs to be sovereign and fully conformant with the law. If you need to change the law to accommodate a specific provider, then you’ve picked the wrong supplier.”

Computer Weekly contacted the Home Office about the changes to the law enforcement data transfer regime, and UK policing’s track record of non-compliance with existing data rules via its use of hyperscalers.

A Home Office source told Computer Weekly that the use of cloud providers, in particular, has caused some confusion, and that measures contained within the bill are intended to give law enforcement the confidence to use cloud processors. However, they said the use of cloud services must not come at the expense of security, and high standards of protection will continue to be applied.

‘Systemic’ transfer issues

Clement-Jones highlighted how cloud service providers routinely process data outside the UK and are unable to provide necessary contractual guarantees to policing bodies, as required by Part Three. “As a result, their use for law enforcement data processing is, on the face of it, not lawful,” he told the House of Lords.

He added this non-compliance creates significant financial exposure for the UK, including potential compensation claims from data subjects for distress or loss, something that is exacerbated by the sheer volume of data pressed by law enforcement bodies: “If only a small percentage of cases result in claims, the compensation burden could reach hundreds of millions of pounds annually.”

Clement-Jones concluded that the government’s attempts to change the law suggest that past processing on cloud service providers has not been compliant with the relevant data protection laws.

As a result, he proposed an amendment “to bring attention to the fact that there are systemic issues with UK law enforcement’s new use of hyperscaler cloud service providers to process personal data”, which would strictly limit overseas transfers to law enforcement bodies with “a legitimate operating need” – that is,  not cloud service providers.

While the Lords were not invited to take a decision on Clement-Jones’s hyperscaler amendment, government minister Baroness Jones said the DUAB’s “bespoke path for personal data transfers from UK controllers to international processors is crucial … [as] we need to ensure that law enforcement can make effective use of them to tackle crime and keep citizens safe”.

One of the biggest problems in data protection is a lack of understanding and clarity [so] anything that can make it clearer and easier to follow can only be a good fit Thomas Barrett, CyXcel

She added the aim of the DUAB’s reform around international law enforcement transfers “is to provide legal clarity in the bill to law enforcement agencies in the UK so that they can embrace the technology they need and make use of international processors with confidence”.

She added: “Such transfers are already permissible under the legislation, but we know that there is some ambiguity in how the law can be applied in practice. This reform intends to remove those obstacles. The noble Lord would like to refrain from divergence from EU law. I believe that in this bill we have drafted the provisions, including this one, with retaining adequacy in mind.”

Barrett said the DUAB will clarify the law in ways that make it easier to put in place contractual provisions and other measures that adequately protect the data: “One of the biggest problems in data protection generally, but particularly here, is a lack of understanding and a lack of clarity … anything that can make it clearer and easier to follow for individuals that have to apply this stuff can only be a good fit.”

Sayers made a similar argument, noting that while many data protection practitioners believe the EU or UK GDPR to be the gold standard of legislation, they “simply fail to recognise that GDPR has a sister piece of legislation in the LED that is sufficiently different that you cannot apply GDPR thinking to it”.

He added: “This is a problem I see day in, day out, where a GDPR hammer is used to try to fix an LED nail, and even the ICO is not immune to confusing the two different sets of laws.”

According to delli Santi, the approach to transfers under the DUAB as it stands is “formalising an approach that has already been changed”. He added that given the deep commercial, governmental and cultural ties between the UK and EU, “the impact of divergence is amplified significantly”. 

Police data logging requirements

The DUAB as introduced will also seek to remove the statutory logging requirements of Part Three, which would allow police to access personal data from various police databases during investigations, without having to manually record the “justification” for the search.

The removal of police logging requirements, however, could represent a further divergence from the EU’s LED, which requires logs to be kept detailing how data is accessed and used.

“The logs of consultation and disclosure shall make it possible to establish the justification, date and time of such operations and, as far as possible, the identification of the person who consulted or disclosed personal data, and the identity of the recipients of such personal data,” says the LED.

Clement-Jones told Computer Weekly that if the law changes to allow police data transfers to, and processing in, infrastructure not owned or controlled by UK bodies, it could “absolutely” be a problem for the UK’s LED adequacy retention. He added that given these clear access and control issues, the potential removal of police logging requirements is “egregious”.

Computer Weekly contacted DSIT about the removal of the logging requirements and whether it believes this measure represents a risk to the UK being able to renew its LED adequacy decision in April 2025, but DSIT declined to comment on the record.

Speaking during the 16 December Lords debate on the bill against the removal of justification logging requirements, Clement-Jones said: “The public needs more, not less, transparency and accountability over how, why and when police staff and officers access and use records about them.”

He added that while policing systems typically capture when, how and by whom data has been accessed, they “very rarely” capture the justification. This is despite the fact that Article 63 of the LED provided a grace period from May 2018 to May 2023 for member states to implement justification recording mechanisms to bring their legacy systems into compliance with the directive – new systems procured from May 2016 onward were required to comply from the start.

To alleviate the issue, Clement-Jones tabled a further amendment to ensure the logging requirements remain, which would “prevent material divergence from the EU Law Enforcement Directive”; although this was also withdrawn.

He also highlighted that “many commodity IT solutions” procured by policing organisations do not capture justifications by default, noting that while a “transitional relief” period was put in place with the introduction of DPA 2018 to modify legacy systems installed before May 2016 – later extended to May 2023 – UK law enforcement bodies did not in general make the required changes.

“Nor, it seems, did it ensure that all IT systems procured after 6 May 2016 included a strict requirement for LED-aligned logging. By adopting and using commodity and hyperscaler cloud services, it has exacerbated this problem,” he said, noting the government now wishes to strike the justification requirements completely.

“This is a serious legislative issue on two counts: it removes important evidence that may identify whether a person was acting with malicious intent when accessing data, as well as removing any deterrent effect of them having to do so; and it directly deviates from a core part of the law enforcement directive and will clearly have an impact on UK data adequacy.”

DSIT claims that removing the logging obligation will save 1.5 million police officer hours a year and save £42.5m for the public purse, but Sayers pointed out that the published impact assessments don’t so far evidence these claims.

“The reality is that most police IT systems don’t have the means to capture the required data,” said Sayers, who was previously involved in the design and delivery of many UK national police systems.

“The factsheets identify this technology problem, which exists on cloud as well as legacy systems like the PNC [Police National Computer], but instead of addressing the issue the government simply want to strike the difficult bits out of the act.”

He added: “The real reason they don’t want to capture the information is they’ve failed to invest any money in upgrading the legacy IT, and the new systems they’ve adopted don’t capture that information by default – and can’t be made to do so.”

DSIT claims that capturing “justification is likely to be of little use in a misconduct investigation”, but Sayers poured cold water on this.

“Public trust, the safety of vulnerable people, as well as the protection of police staff from claims of improper conduct, all rest on being able to prove that access to data was legitimate,” he said.

Home Office figures show police staff misuse of data to be a significant issue, with 1,630 recorded cases investigated in the year to March 2023, the last figures available.

However, Barrett said the removal of justification logging is not a problem, adding it’s more important to have the ability to track who accessed data and when, “because if you’re a bad actor you’re not going to put down the real reason … if you’ve already got access to these kinds of systems, you’re not an idiot, and so you’re going to put something like ‘routine checks’ or some other bland, uninteresting, non-determinative thing”.

He further added that inputting justifications only increases the administrative burden on police, and that while it is very common, even in much older computer systems, to be able to log time and dates, many systems are simply not architected to record justification.

He added: “We’d be much better off making sure that all the systems are really good at recording time and access, because the reality is, in your investigation, that’s going to be the thing that you’re looking at. Not whatever fanciful thing a bad actor has decided to enter as the fake justification for the access.”

During the DUAB debate, Baroness Jones insisted the removal of logging requirements “is not a watering down of provisions. We are just making sure that the safeguards are more appropriate for the sort of abuse that we think might happen in future from police misusing their records.”

While the DUAB has since progressed to readings in the House of Commons, the police data issues were not addressed – outside of vague references to reducing the administrative burden on police officers. It is currently in the committee stage, which will be followed by the report stage and a third reading.

So far, the police data issues have not been discussed during the committee stage.

Source

Posted on

Apple devices are at ‘most risk’ in UK following government ‘backdoor’ order

Users of Apple devices in the UK are “at the most risk in the world” of being hacked, following a secret government order requiring the tech company to allow ‘backdoor’ access to its users’ encrypted data, the House of Lords heard on Monday 31 March.

Liberal peer Paul Strasburger pressed the government to answer questions about a decision by the home secretary, Yvette Cooper, to issue a secret notice against Apple.

The order, first reported in the Wall Street Journal, extends law enforcement and intelligence services’ access to encrypted data stored on Apple’s iCloud to include users of Apple’s secure Advanced Data Protection (ADP) service.

In questions posed in the House of Lords on Monday, Strasburger said the government had “demonstrated its disdain for the privacy and digital security of British citizens and companies” by issuing the TCN against Apple.

The Liberal peer said the order would introduce weaknesses to encryption on Apple devices that could be exploited by criminals and hostile states.

“Strong encryption is essential to protect our data and our commerce from attack by organised crime and rogue states,” he said. “Any weakness inserted into encryption for the benefit of the authorities is also available to those who would do us harm – yet that is precisely what the government are demanding from Apple.”

Tribunal held closed-door hearing

Apple is challenging the legality of the government’s order in the Investigatory Powers Tribunal (IPT), which discussed arguments in a closed-door hearing on 14 March.

Civil society groups Privacy International and Liberty, along with two individuals whose security has been impacted by the government’s order against Apple, have filed separate legal interventions.

Ten newspapers, publishers and broadcasters – including Computer Weekly – have also filed legal submissions calling for Apple’s appeal against the widely publicised order to be heard in open court on public interest grounds.

Non-affiliated peer Claire Fox said it was not possible for Apple to open doors to its customers’ data in a way that would ensure that only the police and intelligence services would have access to its users’ encrypted data.

“It is obvious that criminals, foreign adversaries and others would exploit that weakness,” she said.

Fox said it was baffling if the Home Office was choosing to “bully tech companies into undermining their users’ privacy, security, civil liberties and free speech” while at the same time seeking to establish the UK as a leading hub for innovation and technology.

Liberal democrat peer Tim Clement Jones told the Lords that the government could be in breach of the European Court of Human Rights following a key judgment by the court last year.

In the case of Podchasov v Russia, the European Court of Human Rights found that weakening end-to-end encryption or creating backdoors could not be justified under human rights law.

Labour peer Toby Harris asked what consideration had been given to the trade-off between the “general weakening of security and confidentiality” compared with the gains made by the security services in being able to decrypt data stored by Apple.

Home Office minister and Labour peer David Hanson repeatedly declined to answer questions from peers, citing national security reasons.

“We have a long-standing position of protecting privacy while ensuring that action can be taken against child sexual abusers and terrorists,” he said.

“I cannot comment on operational matters today, including neither confirming nor denying the existence of any notices. This has been the long-standing position of successive UK governments for reasons of national security.”

Conservative peer Daniel Moylan pressed Hanson to comment on Apple’s decision to publicly withdraw its ADP encryption service from the UK, even if he could not comment on whether a notice had been issued.

He also asked the home office minister whether the US and UK governments had any high-level discussions about the order against Apple.  

Bloomberg reported on 13 March that the US and UK governments were holding private talks in an attempt to resolve US concerns that the UK was trying to force Apple to create a backdoor that would allow the UK access to encrypted data belonging to US citizens.

Hanson said he could not comment on the matter.

“Decisions made by Apple are a matter for Apple, and the removal of any features is a matter for Apple. Again, for reasons of national security I cannot confirm or deny any conversations that we have had or any issues that are undertaken,” he said

The Investigatory Powers Act contained “robust safeguards” and “oversight to protect privacy and ensure that data is obtained only on an exceptional basis and only when necessary and proportionate to do so”, he added.

A Home Office spokesperson said: “We do not comment on operational matters, including, for example, confirming or denying the existence of any such notices.”

Media companies have asked the Investigatory Powers Tribunal to hold hearings into Apple’s appeal against the technical capability notice in open court.

Separately, Big Brother Watch, Index on Censorship and the Open Rights Group have written an open letter to the tribunal calling for an open court hearing.

The media companies challenging the secrecy of Apple’s appeal in the Investigatory Powers Tribunal are Associated Newspapers Ltd, the British Broadcasting Corporation, Computer Weekly, Financial Times Group, Guardian News & Media, News Group Newspapers, Reuters News and Media, Sky News, Telegraph Media Group and Times Media.

Source

Posted on

Secret London tribunal to hear appeal in Apple vs government battle over encryption

A secret tribunal is due to meet at the High Court in London this week to hear tech giant Apple appeal against a Home Office order to compromise the encryption of data stored by its customers on the iCloud service worldwide.

The Investigatory Powers Tribunal (IPT) has taken the unusual step of publishing a notification of a closed-door hearing on Friday 14 March, days after leaks revealed that Apple was intending to appeal against the secret order.

Press and civil society groups are expected to petition the Tribunal, which rules on matters of national security, to hold the hearings in open court, given the important public interest surrounding the case and the fact the government’s order has been widely leaked.

The decision by home secretary Yvette Cooper to issue a Technical Capability Notice requiring Apple to give UK law enforcement and intelligence services “backdoor” access to data stored by Apple’s customers on the encrypted version of its iCloud service, has raised tensions between the UK and the US.

US lawmakers are expected to intervene further in the case after the US director of national intelligence Tulsi Gabbard – President Trump’s most senior advisor on intelligence and security – warned that any order from the UK that could put Americans’ privacy at risk would be a “clear and egregious violation”.

As a result of the UK government’s move, Apple in the UK has withdrawn its Advanced Data Protection (ADP) service which allows users to store data in end-to-end encrypted form on iCloud.

The decision is likely to expose people in the UK using Apple services to greater risk of cyber threat as they will no longer have the ability to encrypt their personal data on Apple’s iCloud with end-to-end encryption, though the service will remain available elsewhere in the world.

The president of the IPT, Lord Justice Rabinder Singh, and a senior High Court Judge, Mr Justice Jeremy Johnson, have made themselves available at short notice to hear a case behind closed doors on the morning of 14 March, according to court listings.

The IPT hears national security cases in secure courts at the High Court in the Strand – the only central London venue authorised for national security cases, aside from a secure court on Chancery Lane used for immigration cases.

A series of leaks about the secret order issued by the UK have made it more difficult for the Home Office and security agencies to maintain a stance of neither confirming nor denying the move against Apple.

Privacy International, which has brought a number of cases against government agencies in the IPT, said the Apple hearings should be conducted in public.

Caroline Wilson Palow, legal director and general counsel at Privacy International said: “This is a very important debate to have in public, because we’re talking about the security of our computer systems that can affect millions, if not billions, of people around the world, given the reported technical capability notice has global reach.”

Last month, over 100 cyber security experts, companies and civil society groups signed a letter calling for home secretary Cooper to drop the demands for Apple to create a backdoor that would allow government access to encrypted communications and data stored on Apple’s iCloud service.

Apple has previously said that despite withdrawing Advanced Data Protection from the UK 14 categories of data stored on Apples iCloud will still be end-to-end encrypted by default, including health data.

UK users will not be able to opt for more secure end-to-end encryption for iCloud Backup; iCloud Drive; Photos; Notes; Reminders; Safari Bookmarks; Siri Shortcuts; Voice Memos; Wallet Passes; and Freeform, a collaboration tool.

Source

Posted on

Siri needs its iCloud moment: A complete rebrand

Even without the current Apple Intelligence fiasco, we already realized that Apple needs to ditch Siri or rebrand it in favor of a new personal assistant. Long before Apple Intelligence or LLMs started taking over the internet, we already felt like Siri was lost in time.

In the past few years, reports suggested the issue with Cupertino’s personal assistant is bigger than it seems. In 2023, the New York Times reported about the rise and fall of the assistants, including why Siri struggles with what sounds like regular tasks. John Burkey, a former Apple engineer who worked on the virtual assistant, said it had a “cumbersome design that made it time-consuming to add new features.”

In 2014, he was given the job of improving Siri. But since its database contains a gigantic list of words in nearly two dozen languages, its vast knowledge bade it “one big snowball,” as if someone wants to add a word to Siri’s database, “it goes in one big pile.”

With that in mind, Burkey explained that what seemed like small updates, such as new phrases, would require rebuilding the entire database, which could take up to six weeks. More complex features like new search tools could take nearly a year, meaning Siri could never become a creative assistant like ChatGPT unless it’s completely rebuilt.

Tech. Entertainment. Science. Your inbox.

Sign up for the most interesting tech & entertainment news out there.

By signing up, I agree to the Terms of Use and have reviewed the Privacy Notice.

When looking back at this report, it makes sense why Apple decided to indefinitely delay Siri’s on-screen awareness capabilities, as it still doesn’t know which month we are. This is why I think Apple should rebrand Siri and do something similar to what the MobileMe-iCloud transition was.

Siri feels like MobileMe, but Apple was fast enough to address it

Before iCloud was a thing, Apple had MobileMe. The service was available from July 2008 until October 2011, when iCloud was introduced. However, this subscription-based service was very unstable and had several syncing issues.

This is why when Steve Jobs introduced iCloud and said the service “just worked,” he rhetorically asked: “Why should I believe them? They’re the ones that brought me MobileMe!” Still, iCloud was better than MobileMe, and even though it had a few issues over the years, it’s Apple’s main service.

With that in mind, I think Siri needed a similar approach. At this moment, Apple is focusing on reshaping the personal assistant’s command structure. However, I don’t think promoting executives is enough. If Apple wants to be serious about AI, it must catch up with major players and offer a different experience. Should it call the new assistant Newton, Siri 2.0, or Apple Assistant? It doesn’t matter.

Siri’s revamp is urgent, and Cupertino needs to offer a faster response and service to users.

Source

Posted on

Salesforce execs at TDX 25: Agentforce a whole system AI play

At the TDX 2025 developer conference in San Francisco, Salesforce executives presented its Agentforce agentic AI technology as a “whole system” approach, where large language models (LLMs) are less significant than a “trinity” of data, applications and agents. Relatedly, they consistently disparage “DIY” artificial intelligence (AI) programmes.

Paula Goldman, the supplier’s chief ethical and humane use officer, said: “I think a lot of the public discourse about AI has been about [large language] models. But if you think about Agentforce, it’s a whole system. There’s a foundation model, and then there’s a series of smaller models that go into our Atlas system, and there are workflows that are automated that people can draw on. We’ve got used to talking about AI as models over the past few years, but I think we need to be talking about systems.”

David Schmaier, president and chief product officer at Salesforce, said the supplier’s entire technology stack, including Slack and Tableau, comes into play with Agentforce. He also pointed to its Data Cloud platform as central to its AI offer.

“You couldn’t have a computer without a microprocessor; you need storage and RAM and a display and an operating system around it. That’s what we’ve done. We have our data cloud, which harmonises hundreds of thousands of systems. It gives you the data, the metadata and the semantics. That’s why we can outperform an LLM by itself. LLMs have hallucinations, they have bias, toxicity. An LLM is necessary but insufficient. We add to the LLM. Our view is the data powers the AI and then the AI powers the customer experience of the future,” he said.

An LLM is necessary but insufficient. We add to the LLM. Our view is the data powers the AI and then the AI powers the customer experience of the future David Schmaier, Salesforce

“We call it the ‘holy trinity’. We have the Data Cloud, then we have our Sales Cloud, Service Cloud and Marketing Cloud apps – which is how we got the name Salesforce – as well as Slack, Mulesoft and Tableau. And now we have Agentforce on top of all that. That’s how we can turn on 10,600 customers over three days with agents. It’s because we are using the same platform as we have for 25 years. So, with a healthcare company, for example, that has workflows it has bult in its Salesforce deployment, it can make all those available for [virtual] agents,” Schmaier added.

He believes too many organisations are doing DIY AI. “Most people are just trying to take whatever apps they have, whether it’s Salesforce or SAP or Workday, and just buying ChatGPT and trying to plug it in. No other competitor has what we have, in terms of agents. We think we have a real lead in this agentic field. We’ve sold to 5,200 customers since launching at Dreamforce [in September 2024]. Now, we have 200,000 customers, and most don’t use Agentforce today,” he said.

Rahul Auradkar, executive vice-president and general manager of Unified Data Services and Einstein at Salesforce, made a similar argument about what the provider calls DIY AI.

“What we are doing with agents is an entire system. We’re not shipping a model, an app or a copilot. We’re shipping an AI system on a deeply unified platform. What that system allows our enterprise customers, who don’t want to do the DIY, to do is surface customer-centric analytics and workflows, and listen to the customers to feed back to the system so the agents get better. Copilots are a narrow sliver of what AI can be,” he said.

“The difference between a DIY AI and an enterprise using [our] system is that the enterprise can focus on things that they are good at, which is plenty of things. They have their data. The have their transactions. They have their engagement data. They have their AI policies, their workflows, their automations. We bring all that together within a deeply unified platform and drive value for our customers,” added Auradkar.

DIY AI programmes strongly in evidence among users

And yet, analyst research from Informa TechTarget’s Enterprise Strategy Group (ESG) offers a contrast with Salesforce’s disparagement of DIY AI – a complicating contrast rather than a confutation, but a contrast nevertheless.

Towards the end of 2024, ESG surveyed 832 professionals at organisations across the globe involved in the strategy, decision-making, selection, deployment and management of generative AI (GenAI) initiatives and projects at their organisations and familiar with their organisation’s use of third parties to support GenAI initiatives.

The resulting report, The state of the generative AI market: Widespread transformation continues – authored by Mark Beccue, principal analyst, Mike Leone, practice director and principal analyst, and Emily Marsh, associate research director – does find support for an agentic AI philosophy: “Respondents most often said that they see AI agents, virtual assistants, and intelligent chatbots powered by AI as valuable productivity tools, though they also often said they view them with cautious optimism (41%). Over two-thirds of organisations are planning for or considering AI agents, which represents a significant opportunity for AI vendors to target these requirements with capabilities and services.”

They also note, however: “The AI agent market is extremely nascent and loaded with challenges, including managing single-task agents, interoperability problems, the potential emergence of multitask agents and security.”

But the authors also remark, similarly to Salesforce’s Auradkar, that: “A wide majority (84%) of respondents agreed it is important to incorporate their own enterprise data into models that support generative AI. GenAI models themselves are not a competitive differentiator. Rather, effectively identifying, organising and vetting internal data for use with GenAI models is the key to creating unique and highly actionable insights.”

The research also found user organisations to be embracing a variety of LLMs – open source and proprietary. The largest percentage of respondent organisations (43%) are both proprietary and open source models.

Alongside this enthusiasm for using large language models, the study found that organisations are placing “their bets on internal resources, planning to reskill or upskill employees (58%) and provide education and awareness training to employees (43%)”. This suggests a growing cadre of employees who will want to do DIY AI.

The authors comment: “Employee enthusiasm for these technologies is likely at a high point as GenAI excitement pervades many facets of society, so this internal investment will likely be a win-win situation whereby personnel receive welcome development opportunities and the business gains valuable GenAI expertise.”

At Dreamforce in September 2024, Marc Benioff, co-founder, chairman and CEO of Salesforce, was in combative mood in respect of Agentforce, positioning it as a wholescale alternative to generative AI copilot usage, associated with Microsoft and Google, but with other vendors too.

“There’s a lot of narratives out there from vendors, and a lot of it is not true,” he said at the time. “You need to sit with those customers [at the Dreamforce event], look at the code and break the hypnosis coming from all the vendors. There’s plenty of real customers here who are really deploying real AI. But there are billions being invested in copilots, delivering how much productivity increase? Is there a better way to do it? And so, that’s our gambit.”

The game is still being played. The middle game lies ahead.

Source

Posted on

US Congress demands UK lifts gag on Apple encryption order

US lawmakers have hit out at the Home Office for “attempting to gag” US companies by preventing them from telling Congress whether they have been subject to secret UK orders requiring them to hand over their users’ data.

In an unprecedented intervention, five lawmakers from both sides of the US political divide, led by senator Ron Wyden, have written to the UK’s Investigatory Powers Tribunal (IPT) accusing the British government of undermining Congressional oversight and restricting the free speech of US companies.

Their letter comes as the IPT is preparing to hear closed-door arguments from Apple, which is challenging a notice requiring it to extend UK law enforcement’s existing access to encrypted data stored by customers on the Apple iCloud service anywhere in the world to users of Apple’s Advanced Data Protection (ADP) who choose to hold encryption keys privately on their own devices.

British media organisations, including the BBC, The Times, Financial Times, Reuters, The Guardian, The Telegraph and Computer Weekly, have also filed legal submissions with the IPT today, arguing that there is an important public interest in hearing arguments over the UK’s demands against Apple in a public court.

In the Congressional letter, five US senators and congressmen complained to the Investigatory Powers Tribunal that the secrecy surrounding the orders – known as Technical Capability Notices (TCNs) – are impairing Congress’s power and duty to conduct oversight on matters of national security.

The letter disclosed that Apple and Google have informed Congress that were they to have received Technical Capability Notices, they would be barred by UK law from disclosing it to US lawmakers. The UK embassy has also failed to respond to US requests about potential demands by the UK to other US companies.

“By attempting to gag US companies and prohibit them from answering questions from Congress, the UK is both violating the free speech rights of US companies and impairing Congress’s power and duty to conduct oversight on matters of national security,” the lawmakers wrote.

“The UK’s attempted gag has already restricted US companies from engaging in speech that is constitutionally protected under US law and necessary for ongoing Congressional oversight,” they added.

The letter has been signed by democrats senator Ron Wyden from Oregon, who has campaigned for healthcare and the environment; Alex Padilla from California, who is chairman of the Senate Judiciary Subcommittee on Immigration; and Zoe Loefgren, an advocate for digital rights from California.

By attempting to gag US companies and prohibit them from answering questions from Congress, the UK is both violating the free speech rights of US companies and impairing Congress’s power and duty to conduct oversight on matters of national security Congressional letter to the Investigatory Powers Tribunal

Republicans Andy Bigg from Arizona, chair of the House Judiciary Subcommittee on Crime and Federal Government Surveillance and a vocal trump supporter; and Warren Davidson for Ohio, a member of House Financial Services Committee and a former US soldier, have also signed.

Their unified complaint calls on the IPT to apply principles of open justice to the hearing scheduled for Friday, and for all subsequent proceedings in Apple’s appeal against the Technical Capability Notice. 

The lawmakers note that the existence of the TCN has been widely reported and commented on, which makes any argument for closed hearings to keep the existence of the notice secret “unsustainable”.

The existence of the notice has also been confirmed by Apple’s public decision to withdraw its advanced encryption option, known as Advanced Data Protection, for all UK users. Apple would not have done this “unless it felt compelled to do so by a request to insert a backdoor”.

Holding public hearings would allow lawmakers to hear expert evidence from cyber security specialists, civil society representatives and experts on US-UK data flows, enabling the IPT to reach a well-informed decision over the lawfulness of the notice, they said.

Serious concerns over national security 

The lawmakers argue that the UK’s demands against Apple raise “serious concerns which directly impact national security” and therefore warrant public debate. 

As Computer Weekly previously reported, Tulsi Gabbard, the director of national intelligence, stated in a letter to Congress that the UK’s demands would be “a clear and egregious violation of American’s privacy and civil liberties, and open up a serious vulnerability for cyber exploitation by adversarial actors”.

President Donald Trump confirmed in an interview with The Spectator that he had raised the Apple TCN with prime minister Keir Starmer during his visit to Washington, comparing the UK’s actions to the conduct of China.

Chinese exploited US ‘lawful access’

The lawmakers point out that the security of US technology products against surveillance by foreign governments is an important topic for ongoing Congressional oversight following a spate of hacks against the communications of senior US government officials.

China exploited US lawful interception systems in 2023 to reportedly tap the phone calls of Trump and vice-president JD Vance, and to steal millions of phone records after gaining access to major US carriers in the “Salt Typhoon” attack.

In April 2024, hackers stole phone records of “nearly all” AT&T customers, including records of members of the president’s family, the then vice-president, Kamala Harris, and the wife of the now secretary of state, Marco Rubio, in the “snowflake” incident.

And in 2003, China stole more than 60,000 emails from the department of state and compromised the email accounts of US officials and politicians after hacking into Microsoft-hosted US government email accounts.

“The common link between these incidents is that sensitive government data held by third-party companies was not properly secured and subsequently accessed by hackers … most importantly, the Salt Typhoon incident reportedly involved compromising ‘lawful intercept’ systems of the kind that it appears Apple has been ordered to build,” the letter states.

“Given the significant technical complexity of this issue, as well as the important national security harms that will result from weakening cyber security defences, it is imperative that the UK’s technical demands of Apple – and of any other US companies – be subjected to robust, public analysis and debate by cyber security,” the lawmakers wrote.

Vital for US cyber security experts to comment

“Secret court hearings featuring intelligence agencies and a handful of individuals approved by them do not enable robust challenges on highly technical matters. Moreover, given the potential impact on US national security, it is vital that American cyber security experts be permitted to analyse and comment on the security of what is proposed.”

The Home Office’s shocking order to Apple to break encryption represents a huge attack on privacy rights and is unprecedented in any democracy Rebecca Vincent, Big Brother Watch

The lawmakers invited the tribunal to permit US companies to discuss the technical demands they have received under the UK’s Investigatory Powers Act with Congress. The IPT should “invite robust public debate by independent cyber security experts before deciding the merits of the reported challenge that Apple has brought”, they said.

Separately, civil society groups Big Brother Watch, Index on Censorship and Open Rights Group have written to the president of the Investigatory Powers Tribunal, the Rt Hon Lord Justice Singh, calling for the case to be made public.

They argue that the case implicates the privacy rights of millions of British citizens who use Apple’s technology, and those of its overseas customers.

There is a “significant public interest in knowing when and on what basis the UK government believes that it can compel a private company to undermine the privacy and security of its customers”, according to the letter.

Big Brother Watch interim director Rebecca Vincent said the tribunal hearing must not take place in secret. “The Home Office’s shocking order to Apple to break encryption represents a huge attack on privacy rights and is unprecedented in any democracy,” she said.

Index on Censorship CEO Jemimah Steinfeld said breaking encryption would do away with our rights to privacy, make us far less safe and secure online, and challenge the very notion of the UK as a democracy. “With such high stakes, we demand to know what could possibly justify this. We need answers, not more secrecy,” she said.

Open Rights Group executive director Jim Killock said: “If the UK wants to claim the right to make all of Apple’s users more likely to be hacked and blackmailed, then they should argue for that in an open court.”

Source

Posted on

MSP cuts costs with Scality pay-as-you-go anti-ransomware storage

London-based managed service provider (MSP) Autodata Products has opted for Scality Artesca object storage through its Scality cloud service provider (SCSP) pay-as-you-go purchasing option, which it uses to supply on-premise backup against ransomware for customers.

Benefits of the SCSP licensing model include being able to offer customers highly scalable backup with short recovery time objectives (RTOs) and at the same cost per terabyte (TB) whether it’s for 25TB or 2.5PB (petabytes).

Autodata Products provides IT solutions focused on backup, storage and security via its Cloudlake offer, predominantly based on Wasabi cloud, Veeam backup and Scality storage. It has around 500 customers on rolling monthly contracts and has offices in the US and the Netherlands.

Within its core offer it has Cloudlake Ransomware Recovery Vault (RRV), and it is here that it decided to offer services using Scality Artesca and SCSP. It was already a customer of Veeam’s pay-as-you-go programme.  

RRV is based around the provision of on-site immutable storage for customers. Here, Autodata deploys Scality Artesca object storage as a backup target and pays only for what is used by its customers.

Artesca is Scality’s object storage product aimed at single application use cases and is heavily targeted at data protection. Scality version 3.0 launched in 2024 and emphasised the anti-ransomware resilience capabilities of its object storage platform. These centre on native S3 immutability, anti-exfiltration capabilities, making data indecipherable to attackers at storage and architecture level, as well as by retaining it in multiple geographic instances.  

According to head of datacentre and cloud services Ant Bucknor, Autodata recommends customers keep a workable amount of critical data on-site so they can restore very quickly should a ransomware attack or other outage occur.

He said: “Our clients were restoring their data from the cloud. But that would often break their RTO policy because of the length of time it would take to get everything back up and running, then they would connect to the cloud location and then it would take them longer to bring the data back.”

So, how much data does Autodata recommend customers store on-site?

“I would suggest probably the last 30 days,” said Bucknor. “That would be my base guide, but obviously every client’s different. We’ve got clients where they have data they need to recover quickly from the last six months and others where if it’s over 48 hours old the data is completely worthless.

“The cloud will provide you with a full copy, and it will be immutable. But it isn’t necessarily going to be quick enough.” 

Key to the benefits for Autodata are that it can supply ransomware recovery solutions that would have been out of reach of SME and mid-market customers previously, and that as it buys more product from Scality prices should decrease.

Bucknor said: “Traditionally, these solutions were in the hundreds of thousands of pounds. Whereas, because of the flexibility we have with Scality, we now have solutions that are suitable for SMB, mid-market, education, local government, etc, whereas these solutions just wouldn’t have been accessible to that market before.

“There’s a benefit from a profitability at scale point of view, as in the more of these we do over time, the bigger the benefit there is to Autodata as a business, with a knock-on effect in the better commercial terms for our customers.”

Pay-as-you-go is relatively new in storage purchasing, but it’s a rising trend. HPE offers pay-as-you-go storage as part of its Greenlake offer that stretches across its IT portfolio. NetApp, meanwhile, offers Keystone storage as-a-service, while Pure Storage has its Evergreen storage programmes.

“Pay-as-you-go is the future,” said Bucknor. “The reason is, people want to have a cloud-like purchasing model where they can buy what they want for as long as they want it, and when they don’t want it any more, they can stop paying for it. They want to know what their costs are. Not have bought something over five years and suddenly they want to buy an extra few terabytes of data and it’s three times the price because they’re locked in. People want a more flexible solution.”

Source

Posted on

DeepSeek-R1: Budgeting challenges for on-premise deployments

Until now, IT leaders have needed to consider the cyber security risks posed by allowing users to access large language models (LLMs) like ChatGPT directly via the cloud. The alternative has been to use open source LLMs that can be hosted on-premise or accessed via a private cloud. 

The artificial intelligence (AI) model needs to run in-memory and, when using graphics processing units (GPUs) for AI acceleration, this means IT leaders need to consider the costs associated with purchasing banks of GPUs to build up enough memory to hold the entire model.

Nvidia’s high-end AI acceleration GPU, the H100, is configured with 80Gbytes of random-access memory (RAM), and its specification shows it’s rated at 350w in terms of energy use.

China’s DeepSeek has been able to demonstrate that its R1 LLM can rival US artificial intelligence without the need to resort to the latest GPU hardware. It does, however, benefit from GPU-based AI acceleration.

Nevertheless, deploying a private version of DeepSeek still requires significant hardware investment. To run the entire DeepSeek-R1 model, which has 671 billion parameters in-memory, requires 768Gbytes of memory. With Nvidia H100 GPUs, which are configured with 80GBytes of video memory card each, 10 would be required to ensure the entire DeepSeek-R1 model can run in-memory. 

IT leaders may well be able to negotiate volume discounts, but the cost of just the AI acceleration hardware to run DeepSeek is around $250,000.

Less powerful GPUs can be used, which may help to reduce this figure. But given current GPU prices, a server capable of running the complete 670 billion-parameter DeepSeek-R1 model in-memory is going to cost over $100,000.

The server could be run on public cloud infrastructure. Azure, for instance, offers access to the Nvidia H100 with 900 GBytes of memory for $27.167 per hour, which, on paper, should easily be able to run the 671 billion-parameter DeepSeek-R1 model entirely in-memory.

If this model is used every working day, and assuming a 35-hour week and four weeks a year of holidays and downtime, the annual Azure bill would be almost $46,000 a year. Again, this figure could be reduced significantly to $16.63 per hour ($23,000) per year if there is a three-year commitment.

Less powerful GPUs will clearly cost less, but it’s the memory costs that make these prohibitive. For instance, looking at current Google Cloud pricing, the Nvidia T4 GPU is priced at $0.35 per GPU per hour, and is available with up to four GPUs, giving a total of 64 Gbytes of memory for $1.40 per hour, and 12 would be needed to fit the DeepSeek-R1 671 billion-parameter model entirely-in memory, which works out at $16.80 per hour. With a three-year commitment, this figure comes down to $7.68, which works out at just under $13,000 per year.

A cheaper approach

IT leaders can reduce costs further by avoiding expensive GPUs altogether and relying entirely on general-purpose central processing units (CPUs). This setup is really only suitable when DeepSeek-R1 is used purely for AI inference.

A recent tweet from Matthew Carrigan, machine learning engineer at Hugging Face, suggests such a system could be built using two AMD Epyc server processors and 768 Gbytes of fast memory. The system he presented in a series of tweets could be put together for about $6,000.

Responding to comments on the setup, Carrigan said he is able to achieve a processing rate of six to eight tokens per second, depending on the specific processor and memory speed that is installed. It also depends on the length of the natural language query, but his tweet includes a video showing near-real-time querying of DeepSeek-R1 on the hardware he built based on the dual AMD Epyc setup and 768Gbytes of memory.

Carrigan acknowledges that GPUs will win on speed, but they are expensive. In his series of tweets, he points out that the amount of memory installed has a direct impact on performance. This is due to the way DeepSeek “remembers” previous queries to get to answers quicker. The technique is called Key-Value (KV) caching.

“In testing with longer contexts, the KV cache is actually bigger than I realised,” he said, and suggested that the hardware configuration would require 1TBytes of memory instead of 76Gbytes, when huge volumes of text or context is pasted into the DeepSeek-R1 query prompt.

Buying a prebuilt Dell, HPE or Lenovo server to do something similar is likely to be considerably more expensive, depending on the processor and memory configurations specified.

A different way to address memory costs

Among the approaches that can be taken to reduce memory costs is using multiple tiers of memory controlled by a custom chip. This is what California startup SambaNova has done using its SN40L Reconfigurable Dataflow Unit (RDU) and a proprietary dataflow architecture for three-tier memory.

“DeepSeek-R1 is one of the most advanced frontier AI models available, but its full potential has been limited by the inefficiency of GPUs,” said Rodrigo Liang, CEO of SambaNova.

The company, which was founded in 2017 by a group of ex-Sun/Oracle engineers and has an ongoing collaboration with Stanford University’s electrical engineering department, claims the RDU chip collapses the hardware requirements to run DeepSeek-R1 efficiently from 40 racks down to one rack configured with 16 RDUs.

Earlier this month at the Leap 2025 conference in Riyadh, SambaNova signed a deal to introduce Saudi Arabia’s first sovereign LLM-as-a-service cloud platform. Saud AlSheraihi, vice-president of digital solutions at Saudi Telecom Company, said: “This collaboration with SambaNova marks a significant milestone in our journey to empower Saudi enterprises with sovereign AI capabilities. By offering a secure and scalable inferencing-as-a-service platform, we are enabling organisations to unlock the full potential of their data while maintaining complete control.”

This deal with the Saudi Arabian telco provider illustrates how governments need to consider all options when building out sovereign AI capacity. DeepSeek demonstrated that there are alternative approaches that can be just as effective as the tried and tested method of deploying immense and costly arrays of GPUs.

And while it does indeed run better, when GPU-accelerated AI hardware is present, what SambaNova is claiming is that there is also an alternative way to achieve the same performance for running models like DeepSeek-R1 on-premise, in-memory, without the costs of having to acquire GPUs fitted with the memory the model needs.

Source

Posted on

Balancing act: Managing business needs alongside digital transformation and innovation

When building a startup, there is a real balancing act between managing expectations, educating on what’s possible, and identifying the true cost of innovation. CTOs are challenged not only to build functional technology platforms quickly, but to do so as cost effectively as possible.

Startups are often not profitable therefore don’t have a lot of cash to burn, meaning the CTO has to deliver technology solutions to solve their business goals on a limited budget.

Let’s look at a legacy industry like commercial insurance – it’s been undergoing a transformation in recent years. The industry is data and human heavy and is heavily regulated which is why it’s ripe for innovation. It is also playing catch-up to address the needs of many consumers who want a seamless user experience and businesses that want a modern experience – faster, streamlined, digitised, and so on – when dealing with insurance providers. This is particularly true of the on-demand economy.

Leveraging technology

The on-demand economy is characterised by the likes of Taskrabbit, Doordash, Uber, Deliveroo and Amazon Flex. But it’s the likes of hard working on-demand taxi and delivery drivers who are calling for flexible insurance that caters to their very specific needs which enables them to buy comprehensive coverage for when they’re driving, and to switch it off when they’re not.

However, many insurtechs have not adequately met these needs despite their ability to leverage technology more nimbly and effectively than traditional players. The business of insurance is complicated and innovation cannot be retrofitted with existing tech, which is why it’s vital to have a deep understanding of what the requirements are between the customer, the insurance partners and platforms like Uber and Amazon, for instance.

Transforming the on-demand insurance industry is a symbiotic relationship between the customer, the insurance provider and the platform. Although it can deliver real results for all, it also comes with its share of unique challenges.

Loss ratio – how much an insurance company spends on claims compared to the premiums it receives – is a key indicator of profitability. When insurtech startups focus too much on showy AI-driven gimmicks such as automatic claims payments within seconds, loss ratios suffer – and crucial insurance industry partners back away quickly. In the world of insurance, “innovation at all costs” simply doesn’t work.

But technology cannot simply operate as a cost centre. By working in partnership with the rest of the business, startup CTOs and their teams need to focus on building an ongoing technology foundation to drive innovation within legacy industry structures and processes, driving business growth as well as consistent results for customers and partners.

Tech as augmentor – not replacement

Many of the challenges CTOs face aren’t necessarily about technology, but the change of mindset required when implementing tech solutions. Until very recently, insurance was an industry dominated by traditional players, governed by outdated systems and processes. While this is changing, there are still areas where bridges must be built between the promise of what technology can deliver and a certain “this is how it’s always been done” mindset.

For example, we know that insurance, like many industries, is ripe for reinvention through smart uses of AI – as long as it is implemented in the most appropriate areas of the business, and used as an augmented assistant rather than a replacement for specialist expertise.

Chris Gray headshot

“Many of the challenges CTOs face aren’t necessarily about technology, but the change of mindset required when implementing tech solutions”

Chris Gray, Inshur

At Inshur, working in combination with a team from Google Cloud, we were able to build an AI assistant for our claims team and demonstrate to management its effectiveness in helping the team prioritise work as well as speeding up administrative tasks, while providing fast and effective customer service. We’re continuing to roll out this technology internationally, as well as add further features to augment the human adjusters and utilise their expertise while saving them time.

The assistant helps the team to quickly scan incoming documents, including email, physical letters, attachments or transcribed phone calls; infer the data, including who is the sender and the intention of the communication; identify important and useful information such as vehicle registration and claimant name; identify the priority and urgency of the claim; assign it to the right team; and summarise the data into a standard format for ease of use. By automatically accepting feedback, retraining, and learning from past actions, the assistant also helps guide handlers with proposed next steps, helping to train new claims handlers.

The AI-based tools we built to support our claims teams have enabled us to see patterns that are also a good fit for other departments within the business. So much so, that we see potential for the commoditisation of these approaches to a wider set of solutions that serves not just insurance, but any business.

Build or buy?

Another question a lot of startup CTOs are asked is whether to build or buy. Building tech solutions from scratch can carry significant risk, especially given the resource investment typically required. But when every business in a given market is using the same platforms – usually with significant tweaks and workarounds to fit their specific needs – then nobody can truly win the innovation race.

First-movers must always be willing to build when necessary, and to buy when prudent.

For example, we decided that we needed to invest in developing our own solutions to problems that could not be adequately solved by off-the-shelf products. One such product is our Pay-as-you-flex wallet for Amazon Flex. While traditional insurance has historically covered drivers at all times, including when they’re not driving, we knew that technology held the key to delivering a new insurance product that would enable delivery drivers to pay only for the cover they needed, when they needed it.

As the first-of-its-kind to enter the market, we knew that we’d need to build it from scratch.

It’s only since we built our proprietary platform to manage business-critical processes including policy administration, claims management and billing that similar products have entered the market. By building a platform that’s fully tailored to the specific needs of the market we serve, we’ve paved the way for other insurers to do the same for their customers and partners.

However, the startup CTO must also take the lead in conversations where buying makes most sense, securing buy-in from other senior stakeholders and identifying the most appropriate vendors to partner with. Often, particularly in a high-growth startup where cost and return on investment are key considerations, this will involve a detailed assessment of risk for all available scenarios.

In Inshur’s case, we’re working with Google Cloud to implement several of its AI products to drive efficiencies and ensure that customers are treated fairly – which is both a regulatory and moral imperative in the insurance industry.

We know that our customers drive for a living, which means they often need to call us via their hands-free mobile technology while driving in between journeys, rather than emailing or speaking to a text-based chatbot. 

When we identified that a significant proportion of the calls coming into our customer service team could be quickly and effectively answered by an AI-driven solution, we implemented a “smart virtual agent” to handle more straightforward queries, enabling the team to focus more on serving customers with specific or detailed questions.

Bridging the gap

Because of the crucial role technology such as AI will play in the coming years, CTOs will need to ensure they are consistently developing deep understanding and expertise, not just in the latest technology innovations but also how they can be implemented to drive business strategy and growth.

Crucially, this will include taking a leadership role in helping to educate stakeholders across the business on the best use cases for AI tools and other solutions, building understanding at every level around what the technology can and can’t help with, and putting clear structure and process around innovation.

This ability to bridge the gap between the business and technology is already becoming a crucial indicator of future success.

Chris Gray is chief technology officer at vehicle insurance provider Inshur.

Source

Posted on

Gemini AI just got a new feature ChatGPT can’t match yet

The smarter AI programs like ChatGPT and Gemini become, the more we’ll want to use them as the virtual assistants they can be. For that to happen, we’ll need the AIs to access information about us from all sorts of apps and remember details about us. We’ll also need to be able to trust companies like OpenAI and Google with increasingly more personal data.

OpenAI was the first to bring memory features to ChatGPT. It happened with Custom Instructions, a feature I’ve used since it became available. About a year ago, OpenAI also added a Memory feature to ChatGPT that allowed it to remember things about users from chats beyond the scope of Custom Instructions. All of this happens with the user’s knowledge, and memories can be deleted at any time. Also, they don’t train the AI if you set your ChatGPT privacy preferences correctly.

Gemini needed more time to get memory features similar to ChatGPT. Google rolled out the first memory features in November, but they’re available to Gemini Advanced subscribers. ChatGPT Memory features are also available to paying ChatGPT users.

However, Google has now improved Gemini’s memory in a way that OpenAI hasn’t. You can tell Gemini to recall information from your previous chats with the AI on a similar topic, which can be handy for picking up a conversation on the same subject.

Tech. Entertainment. Science. Your inbox.

Sign up for the most interesting tech & entertainment news out there.

By signing up, I agree to the Terms of Use and have reviewed the Privacy Notice.

“Starting today, Gemini can now recall your past chats to provide more helpful responses,” Google said in a blog on Thursday. “Whether you’re asking a question about something you’ve already discussed, or asking Gemini to summarize a previous conversation, Gemini now uses information from relevant chats to craft a response.”

While I have Custom Instructions enabled in ChatGPT and update them from time to time, I’m not using the memory feature. I don’t fully trust the AI to remember information about me, not that I provide information that might be too personal to hand over to the AI to begin with.

However, Google’s upgrade for Gemini is something I’d want from ChatGPT. The ability for ChatGPT to recall some conversations on a similar topic would certainly come in handy, as it would prevent me from having identical chats. That can happen from time to time.

I will remind you that ChatGPT Search did give ChatGPT a major UI overhaul, allowing users to search for previous chats. This makes it somewhat easier to recall past conversations, but I have to do it manually. Also, ChatGPT supports folders, so I can combine similar chats in the same folder to streamline my interactions with the AI.

Google’s way is better. I’d want to tell the AI to look at past conversations and find relevant information. This isn’t necessarily the same thing as the memory feature. It’s just giving the AI access to my chat data already stored in my account with a twist. I’d be able to manage what data the AI sees.

Google says that’s the case with Gemini:

You’re in control over what information is stored. You can easily review, delete or decide how long to keep your chat history. You can also turn off Gemini Apps Activity altogether by going to My Activity. Gemini may indicate when it uses your past chats in sources and related content.

The new memory feature is rolling out in English and you’ll need a Gemini Advanced subscription via the Google One AI Premium Plan. This subscription also gives you access to Google Cloud storage, which makes it a better deal than ChatGPT Plus.

Google Workspace Business and Enterprise subscribers will also get the feature in the coming weeks.

Source