Author: Eternity Lab

The Metropolitan Police’s rapid “unchecked” expansion of live facial-recognition (LFR) technology is taking place without clear legal authority and minimal public accountability, says Green London Assembly member Zoë Garbett in a call for the force to halt its deployments of the controversial technology.

Made during an ongoing government consultation on a legal framework for the technology, Garbett’s call for the force to immediately halt its deployments of LFR is informed by concerns around its disproportionate effects on Black and brown communities, a lack of specific legal powers dictating how police can use the tech, and the Met’s opacity around the true costs of deploying.

Garbett’s intervention also comes as the High Court is considering the lawfulness of the Met’s approach to LFR, and whether it has effective safeguards or constraints in place to protect people’s human rights from the biometric surveillance being conducted.

“Live facial-recognition technology subjects everyone to constant surveillance, which goes against the democratic principle that people should not be monitored unless there is suspicion of wrongdoing,” said Garbett, adding that there have already been instances of “real harm” in children being wrongly placed on watchlists, and the disproportionate targeting and misidentification of Black Londoners.

“These invasive tools allow the police to monitor the daily lives of Londoners, entirely unregulated and without any safeguards. The Met repeatedly claim that live facial recognition is a success, yet they continue to withhold the data required to scrutinise those claims.

“It makes no sense for the home secretary to announce the expansion of live facial recognition at the same time as running a government consultation on the use of this technology. This expansion is especially concerning given that there is still no specific law authorising the use of this technology.”

Highlighting in a corresponding report how facial-recognition technology “flips the presumption of innocence” by turning public spaces into an “identification parade”, Garbett also outlined ways in which both the Met and the Home Office can make its use safer in lieu of a full-blown ban.

This includes creating primary legislation with “strict controls” that limits LFR to the most serious crimes and bans its use by non-law enforcement public authorities or the private sector; and openly publishing deployment assessments so that watchlist creation, location choice and tactical decisions are publicly available for Londoners to review.

On watchlist creation specifically, Garbett dismissed the police claim that LFR is a “precise” tool, highlighting how nearly every watchlist used is larger than the one preceding it.

Highlighting how the number of faces being scanned by the Met is “increasing at a near exponential rate”, Garbett likened the forces watchlist tactics to a “fishing trawler” that it keeps adding to so it can find people.

“Data suggests that rather than making a new unique watchlist for each deployment based on the likelihood of people being in the area of the deployment, it seems from the outside that the MPS is just adding additional people on to a base watchlist [it has],” she said.

Garbett also called on the Met to publish the true financial and operational costs of all LFR deployments, arguing that the force has not only failed to provide a compelling business case for the technology, but is actively obfuscating this information.

“The MPS has a history of a lack of transparency. This is perhaps best summarised by Baroness Casey in her review of the MPS where she said, ‘The Met itself sees scrutiny as an intrusion. This is both short-sighted and unethical. As a public body with powers over the public it needs to be transparent to Londoners for its actions to earn their trust, confidence and respect’,” said Garbett.

She added that while freedom of information requests returned in mid-2023 revealed that, up until that point, the force had spent £500,000 on the tech, without up-to-date reliable figures, it is impossible to verify the Met’s claims that it is delivering a greater impact on public safety through LFR.

“The NHS wouldn’t be able to roll out a new treatment without being able to prove it was worthwhile and effective, but it seems that the police operate under their own rules and seemingly answer to no one,” said Garbett.

Computer Weekly contacted the Met about Garbett’s report. A spokesperson said that LFR “has taken more than 1,700 dangerous offenders off the streets since the start of 2024, including those wanted for serious offences, such as violence against women and girls. This success has meant 85% Londoners support our use of the technology to keep them safe.

“It has been deployed across all 32 boroughs in London, with each use carefully planned to ensure we are deploying to areas where there is the greatest threat to public safety. A hearing into our use of live facial recognition has taken place and we look forward to receiving the High Court’s decision in due course. We remain confident our use of LFR is lawful and follows the policy which is published online.”

A lack of meaningful consultation so far

While the use of LFR by police – beginning with the Met’s deployment at Notting Hill Carnival in August 2016 – has already ramped up massively in recent years, there has so far been minimal public debate or consultation, with the Home Office claiming for years that there is already “comprehensive” legal framework in place.

The lack of meaningful engagement with the public by police and government over facial recognition is reflected in Garbett’s report. She highlights, for examples, that Newham Council unanimously passed a motion in January 2023 to suspend the use of LFR throughout the borough until biometric and anti-discrimination safeguards are in place.

While the motion highlighted the potential of LFR to “exacerbate racist outcomes in policing” – particularly in Newham, the most ethnically diverse of all local authorities in England and Wales – both the Met and the Home Office said that they would press forward with the deployments anyway.

“Since that motion was passed, LFR has been used 31 times in Newham by the MPS,” said Garbett.

On the deployment of permanent LFR cameras mounted to street furniture in Croydon, Garbett added while the Met promised it would consult with the local community, councillors from there are have told her the force did not follow through with this consultation.

The technology was similarly rolled out in Lewisham without meaningful consultation, despite the Met’s claims to the contrary.

However, in December 2025, the Home Office launched a 10-week consultation on the use of LFR by UK police, allowing interested parties and members of the public to share their views on how the controversial technology should be regulated.

The department has said that although a “patchwork” legal framework for police facial recognition exists (including for the increasing use of the retrospective and “operator-initiated” versions of the technology), it does not give police themselves the confidence to “use it at significantly greater scale…nor does it consistently give the public the confidence that it will be used responsibly”.

It added that the current rules governing police LFR use are “complicated and difficult to understand”, and that an ordinary member of the public would be required to read four pieces of legislation, police national guidance documents and a range of detailed legal or data protection documents from individual forces to fully understand the basis for LFR use on their high streets.

Consultation responses

In a section on how people can respond to the Home Office’s LFR consultation, Garbett urged people to call for its ban, adding that further protections in lieu one could include requiring a warrant to be placed on a watchlist, and limiting it to “the most serious and urgent crime purposes”.

She noted that, as it stands, the Met has not used LFR to make any terror-related arrests, with the most common offence being variations on theft or court order breaches

“In a recent press release, the lead example the MPS give for how they have used LFR is using it to arrest a 36-year-old woman who was wanted for failing to appear at court for an assault in 2004 when they were probably 15 years old,” she said. “The public might feel differently about LFR if they knew it was being used on cases such as these.”

On the permanent installation of LFR cameras in Croydon, Garbett added that while the police have said they are only switched on when an operation is taking place, “there is still the potential for 24/7 monitoring, with Londoners unable to tell if the cameras are operational or not. This makes the feeling of being under surveillance in London feel routine and begins to be a slippery slope to preventative policing and a blurry line between safety and social control.”  

Garbett concluded that the rapid deployment of LFR must stop while safeguards are in place to protect people’s rights: “I urge everyone to respond to the government consultation and use the guide I’ve prepared to make sure we have a say in how this technology is used going forward.”

Computer Weekly contacted the Home Office about the contents of Garbett’s report and its decision to massively expand facial-recognition deployments before concluding its consultation.

“Facial recognition is a crucial tool that helps the police locate suspects and those wanted by the courts. In the past two years alone, it has helped the Metropolitan Police locate and arrest more than 1,700 offenders, including rapists, domestic abusers and sex offenders,” said a spokesperson.

“The Home Secretary has announced plans to roll out facial recognition across the country. The number of live facial recognition vans will triple, with 50 vans available to every police force in England and Wales.”

Source

The total number of common vulnerabilities and exposures (CVEs) disclosed in 2026 is set to romp past the 50,000 mark in 2026 and may plausibly run as high as six figures for the first time ever, according to the Forum of Incident Response and Security Teams’ (First’s) annual Vulnerability report.

In its latest set of predictions, First said that this year, the upper bounds of its 90% confidence interval in fact approaches 118,000 CVEs, and according to the data, realistic scenarios suggest 70,000 to 100,000 disclosed vulnerabilities are “entirely possible”. The median figure for 2026, it said, would most likely be around 59,000.

First said that whatever the figure turns out to be, it underscored an “urgent need” for organisations to both scale their security ops and strategically prioritise their vulnerability response and patching practices.

“The question organisations need to ask right now is: are my people and processes ready to handle this volume, and am I prioritising the vulnerabilities that actually put my data at risk?” said Éireann Leverett, first liaison and lead member of First’s Vulnerability Forecasting Team.

“Our forecast allows defenders to stop reacting to every new CVE and start making strategic decisions about where to focus limited resources before attackers exploit the gaps.”

The 50,000 vulnerability question

In its 2025 report, First said that the higher end of its predicted range topped out at 50,000 CVEs – the number its analysts expect to comfortably exceed this year. This was partly due to the rapid adoption of open source software (OSS) and the use of AI tools both in vulnerability discovery. During the course of the year, the emergence of the vibecoding phenomenon likely also had an impact.

In the event, First’s prediction was bang on, Leverett revealed, tipping over the upper confidence mark on 31 December 2025 for a final total of 49,972 observed CVEs, just 28 short of the magic number.

However, ideally, the upper confidence point would fall somewhere in 2026, with the median confidence point falling on New Year’s Eve, and as a result, First has reviewed its approaches and methodology going forward. Whether or not this means its 2026 forecast will be even more accurate remains to be seen.

“[Our] new method of forecasting…allows for asymmetric confidence intervals. This means we are taking into account that the publication number is more likely to exceed last year than be less than last year,” Leverett told Computer Weekly.

“So while we expect the number to be closer to 60,000, there is a 10% chance it exceed 118,000. Most of this is just statistics, but there is also discussion about emerging technologies and how they might stretch the range of possible numbers, which meant we were more comfortable publishing the results of this modelled outcome than some others.”

Next steps

While at first glance First’s annual CVE report might seem just an interesting statistical marker, the forecast serves as a potentially critical planning tool for the security sector when it comes to planning patching capacity, writing coordinated disclosures, or developing new detection signatures for SIEM, EDR or IDS platforms.

“Much like a city planner considering population growth before commissioning new infrastructure, security teams benefit from understanding the likely volume and shape of vulnerabilities they will need to process,” said Leverett. “The difference between preparing for 30,000 vulnerabilities and 100,000 is not merely operational, it’s strategic.”

Whether they end up facing 50,000 or 100,000 CVEs, and keeping in mind that not every flaw will affect every business, security leaders at end-user organisations can start the work to get out in front of the problem right now.

A strong jumping off point is to assess whether the organisation has the people, processes and capacity to handle so many issues. A well-prepared CISO will have prepared for the median forecast but will also have built contingency plans for the higher-volume scenarios.

Security pros also need to master the art of ruthless prioritisation, focusing on the flaws that pose the greatest risk to their specific IT estates, and not just those with the most critical CVSS numbers.

Finally, leaders should leverage external vulnerability forecasts alongside their own asset inventories to make vendor- and product-specific preparations.

“No company can solve vulnerabilities and cyber security in isolation. The organisations that recover fastest are the ones with trusted networks already in place, sharing threat intelligence and coordinating response before a crisis hits,” said First CEO Chris Gibson.

Source

The National Audit Office (NAO) has refused to endorse the audited accounts of the College of Policing for a second year running, as the policing organisation continues to recover from serious failures in an IT project that left it unable to properly manage its finances.

The accounting watchdog said that, although there had been no new financial issues, it “disclaimed” the college’s latest 2024–25 accounts because of the continuing “fallout” from financial reporting problems that resulted from a problematic IT project.

“We were not able to provide a complete opinion on the opening position or in-year transactions for 2024–25, both of which are heavily derived from the closing position of the prior year,” it added.

The College of Policing, the professional body for policing in England and Wales, ran into difficulties when it replaced its SAP-based accounting systems with the Home Office’s Oracle-based Metis accounting system in October 2023.

It transferred its payroll systems from CGI UK IT limited to a new supplier, Shared Services Connected Limited (SCCL), on the same day.

‘Significant concerns’

The move left the college unable to produce accurate figures for financial transactions, leading to “significant concerns” about the integrity of the college’s financial records, which contributed to an overspend of £1.3m.

College CEO Andy Marsh said in a statement that it had now addressed its previous financial problems.

“We have introduced stringent new procedures to stabilise our financial processes and made further progress by enhancing the college’s financial controls, including building greater resilience and expertise at board, executive and operational levels,” he added.

Computer Weekly reported in July 2025 that the NAO found the college failed to manage the risks of the project, and did not address known defects before going live with its new accounting systems.

There was a failure to segregate financial report data held by the Home Office and the college, and the problem remained unresolved during testing and “go-live”, leading to “potential inaccuracies” in financial reporting, the NAO found. Issues with data conversion and migration were also not resolved, creating further risks to the integrity of financial data.

The college had failed to check on a “line by line” basis that the transactions on the SAP systems had been accurately and completely transferred to Metis, an outsourced service shared with other government departments, and was unable to obtain “a significant amount of information” required from its 2023–24 financial audit.

Contract issues

The Home Office’s contact with SCCL did not require the service provider to hand over the payroll information the college needed for its 2023–24 audit, delaying the information required by auditors to complete their work by four months.

The problems were exacerbated because the college lacked people with the right technical and financial skills. It had only one member of staff with knowledge of the SAP accounting system, who went on an extended leave of absence, and the board member overseeing the accounting team was not a qualified accountant.

Marsh added that the National Audit Office acknowledged the improvements the college has made this year.

“The auditor stated that: ‘For 2024 to 2025, the college has successfully produced a set of auditable financial statements, which is a significant achievement from a difficult starting position.’ This progress represents a crucial step in the college’s financial recovery and is a notable achievement given our challenging starting point,” he said.

“In 2023, we encountered major challenges with our accounts, caused by the introduction of a new finance and HR system. The auditor’s disclaimer on the 2024 to 2025 accounts relates solely to these previous financial problems, which have now been addressed.”

The college described this year’s audit as the first step in a three-year audit recovery plan. It remained “on trajectory” to restore a fully unqualified audit opinion in 2026–27, according to its published accounts.

The college said it had conducted a lessons learned exercise and undertaken best practices training with the NAO, in addition to appointing a chief financial officer and director of delivery.

It addressed 40 technical and systems issues identified by the NAO, and is continuing to work through the list with the support of the Home Office.

The college also worked with SSCL to ensure that previous problems with not having “timely access” to audit information – particularly payroll – were not repeated.

Source

The European Commission (EC) has preliminarily found that TikTok’s addictive design features violate the bloc’s Digital Services Act (DSA).

The preliminary decision outlines how addictive design features on the platform, such as infinite scroll and autoplay, are resulting in users going into “autopilot mode”, with the EC stating this may lead to “compulsive use”.

The DSA, which sets out rules for online services used by European citizens, is designed to strengthen consumer rights and consumer choice, while also minimising the risk of harm. The act also requires platforms to carry out a risk assessment of negative effects on children’s mental health and present it to the EC.

TikTok is one of the 17 companies defined as Very Large Online Platforms under the act, which means it has to comply with the most stringent rules of the DSA because the size of its user base means there is greater potential for systemic harms to occur.

In its ruling, the EC stated that TikTok had failed to implement reasonable and effective measures to mitigate risks from its addictive design features, arguing that minors and vulnerable adults are at particular risk of harm.

The EC’s investigation also revealed that TikTok’s risk assessment had not adequately addressed how its design features and dark patterns could cause harm to the physical and mental health of its users.

On the protective measures that are in place, including screen time management and parental control tools, the EC noted they “do not seem to effectively reduce the risks stemming from TikTok’s addictive design” due to being easy to dismiss or overlook.

“At this stage, the Commission considers that TikTok needs to change the basic design of its service. For instance, by disabling key addictive features such as ‘infinite scroll’ over time, implementing effective ‘screen time breaks’, including during the night, and adapting its recommender system,” it said.

European Union (EU) tech chief Henna Vikkunen told reporters that minors are more at risk because “they don’t have the same tools” to avoid compulsive behaviour.

The decision could force the app, which has more than one billion users globally, to make design changes to avoid penalties, the European Commission said.

If it fails to make the necessary changes, the app could face fines of up to 6% of annual revenue, which it reportedly hoped could reach $186bn last year. TikTok was also accused of breaking digital advertising rules over transparency in May 2025.

The ruling marks the first time the EC has taken a legal stance on the design features of a social media company, tackling what many online safety advocates and campaigners recognise as addictive design.

Responding to the EC’s preliminary findings, a TikTok spokesperson said they “present a categorically false and entirely meritless depiction of our platform, and we will take whatever steps are necessary to challenge these findings through every means available to us”.

However, many hope this could set a precedent for future action against recommender systems that amplify allegedly harmful and illegal content. In December, the first-ever fine for breaching the DSA was given to X, with a total of £104m.

These findings confirm what people have been saying for years: TikTok’s addictive design is not an accident, it’s a business model Ava Lee, People vs Big Tech

“These findings confirm what people have been saying for years: TikTok’s addictive design is not an accident, it’s a business model. We need [EC president] Ursula von der Leyen to stand up for European citizens and show the political leadership this moment calls for,” said Ava Lee, executive director of People vs Big Tech.

US states have already pursued a case against TikTok for its addictive design features, with the lawsuit alleging the product is damaging children’s mental health. 

The move by the EC comes at a time when concerns around the safety of social media platforms are growing across the continent.

France, for example, has voted on a social media ban for children under the age of 15, while Spain has proposed to criminalise algorithms that amplify illegal content. In his speech to the World Governance Summit in Dubai, Spanish prime minister Pedro Sanchez vowed to protect children “from the Digital Wild West”.

In January 2026, the UK’s House of Lords voted to back a social media ban for under-16s by 261 votes to 150, with the government launching a national consultation to discuss next steps for online safety and digital well-being.

Von der Leyen has expressed support for an EU-wide age limit, following Australia being the first country in the world to ban under-16s from accessing social media in December last year.

“Amidst current discussions of restrictions on children’s access to social media platforms, governments must remember they also have a duty to protect children’s right to participate in the digital world,” said Lisa Dittmer, Amnesty International researcher on children and young people’s digital rights.

“To do so, their focus must be on tackling the toxic design of leading social media platforms, including through effectively enforcing laws like the Digital Services Act,” she added.

Research by Amnesty International has previously found that despite risk mitigation measures announced by TikTok since 2024, the platform continues to expose vulnerable users to content that normalises self-harm, despair and suicidal ideation.  

The European Commission’s announcement of its preliminary findings comes just days after a report was published by the US government’s House Judiciary Committee, titled The foreign censorship threat, Part II: Europe’s decade-long campaign to censor the global internet and how it harms American speech in the United States.

The publication states that the European Commission has “pressured major social media platforms to change their global content moderation rules”, revealing growing discontent from the Trump administration towards EU tech regulation.

“For the first time, the European Commission is critically examining the recommender mechanisms through which platform operators manipulate the free choice of users,” said German MEP Alexandra Geese. 

“Coupled with high personalisation, this system distorts the idea of freedom online. I expect these recommender mechanisms to be scrutinised on other platforms, too. There are better algorithms for ensuring choice online. It’s not the users who want disinformation, hate and violence, but the platforms.”

Source

Following a report that Apple pushed back its plans to introduce an AI health coach subscription as part of its new features expected for iOS 27, Google is doing the exact opposite, introducing an AI health coach to Fitbit Premium users on iOS. According to a Fitbit community blog post, the health tracker is offering a Public Preview of its upcoming AI personal health coach for iOS users.

For now, only paid or trial Fitbit Premium subscribers with an up-to-date app and operating system, in addition to a few other requirements, can join the Public Preview. If they meet the criteria, they can take advantage of Google’s AI health coach, which has been available to try on Android devices for a few months now. Google says this 24/7 digital advisor can offer tips and guidance across three main points: fitness, sleep, and holistic health, which are all features the activity tracker already offers.

Here’s everything Fitbit’s AI coach will be tackling

According to Fitbit, the experience of using its AI health coach will be like going to the gym for the first time. You’ll talk with the AI about your goals and lifestyle, so it can understand how to provide you with the best data (as if you live an active life, if you’re trying to lose weight, get back on track after an injury, and so on).

The difference of this AI coach is that it’s expected to be highly adaptable, so if you had a poor night’s sleep, there’s something on your schedule, or priorities changed, it will give you guidance based on everything happening in your life. Besides that, Google says this AI coach looks at the “bigger picture,” and not only at daily goals, as the ultimate goal is to bring consistency to users’ health and fitness initiatives.

In an enterprise page, Google also highlights how the personal health coach can be great for preventive care, as it can understand and interpret your wellness data, in addition to overall fitness information and sleep consistency. The company says the health data isn’t used for Google Ads, and that users have total control over what to share with the AI personal coach and also what not. Still, not everybody is sold on these privacy claims.

What happened to Apple AI coach efforts?

José Adorno/BGR

Apple has reportedly been planning to introduce its AI health coach for at least over a year, as this feature was expected for last year’s WWDC 2025. While the company postponed the announcement, recent rumors indicate that it could be unveiled as soon as iOS 26.4, or even at the WWDC 2026 with the iOS 27 introduction.

However, with Bloomberg saying Apple scrapped its plans due to the AI health coach not meeting the standards for a new subscription, it seems the company might slowly integrate some of its ongoing efforts on the Health app in the upcoming updates. For example, Apple has reportedly made a studio in California to record medical-based information about health data, offering users guidance over their data.

Besides that, the company just started testing AI features for Apple Watch, with functions like Workout Buddy, and Apple Fitness+ AI dubbed programs to offer its catalog to a broader audience. At the end of the day, it seems Apple might continue to add value to its current subscriptions and features, even though it isn’t offering anything on top of what competitors already are. That said, we’ll discover more about its health-related plans in a couple of months, once WWDC 2026 kicks in.

Source

Digital twins – virtual representations of environments and dynamics of interest – can address a wide range of decision-making needs and opportunities, and expectations for the technology and related applications are high.

A study from Fortune Business Insights projects the market to grow from $24bn in 2025 to more than $250bn in 2032. Digital twins can support research, planning and operations across a wide range of application areas, such as biological systems, machines and infrastructures, industrial operations, communities and cities, and even simulations of global and geopolitical dynamics. Recent discussions have centred on their use for robotics and robotics management.

The versatility of digital twins is substantial, but hurdles exist that prevent them from reaching their full potential. Some dynamics are not fully captured, while other dynamics are difficult to address comprehensively. In some cases, artificial intelligence (AI) can reduce existing limitations, but use of AI can create its own problematic issues.

Injecting human behaviour

Machines and equipment can be modelled according to physical formulas and accumulated sensor data that capture real-time and real-world behaviour. The same is true for electricity and water networks, for example. These systems are complicated but can be modelled in theory.

Complicated systems behave in predictable ways. Complex systems, in contrast, will behave differently each time, in part because of human behaviour that can change according to many influences. Most digital twins tend to omit human behaviour, while others treat human behaviour as predictable placeholders – in a way, they mechanise human behaviour. But human behaviour and interactions are of crucial importance in simulating dynamics for digital twins for cities and urban environments – after all, that’s what cities and communities are ultimately created for.

Farzin Lotfi-Jam, assistant professor at Cornell University’s College of Architecture, studies the use of technologies to govern cities. He is the director of Cornell University’s Realtime Urbanism Lab, which “investigates the impacts of new technologies that virtualise cities and populations”.

He says: “The global proliferation of urban digital twin models compels a research agenda that investigates the intertwined social, political and technical dimensions of their development, from design to use in planning and governance. In each of these digital twinning concepts is a concept of what a city is. I noticed, looking at all of these, that there’s no people anywhere in any of these concepts.”

A research field far removed from Lotfi-Jam interests could potentially add guidance in populating digital twins for cities. Tianyi Peng, assistant professor in the decision, risk and operations division at Columbia Business School, is looking at the use of AI for decision making. His research looks at what can be used to generate AI agents that mimic human behaviour, such as that in the context of market decisions like shopping preferences and reactions to product stimuli.

The current use of digital twins for urban environments is limited for the lack of realistic representations of humans and their actions and interactions. It is easy to see how the study of individual behaviour and the simulation of group interactions will find use in city digital twins over time.

Peng’s colleague Olivier Toubia, professor of business at Columbia Business School, who is investigating interactions between AI-generated behaviour and how these interactions affect collective behaviour, “combines methods from social sciences and data science to study human processes such as motivation, choice, and creativity”.

Meanwhile, Lydia Chilton, associate professor of computer science at Columbia University School of Engineering & Applied Sciences, is contributing research into how AI agents in simulated environments can mimic unique behaviours of human interactions that can be unpredictable.

Providing comprehensive data

Mutualistic technologies offer ways looks at the wider set of technologies that interact with each other with impact on digital twins and robotics. The emerging network of mutualistic technologies features AI and sensors as the glue that creates positive feedback loops between these technologies. Data is needed to create realistic representations and relevant interactions between virtual and real world. Many times, real-world data can be difficult or expensive to extract though. Then synthetic data can find use. Synesthetic data can come from simulations in digital twins or from AI-based applications.

Commercial relevance of capturing comprehensive data is substantial, particularly for digital twins for urban environments where data from many interdependent networks require inclusion to realistically mirror activities and interactions of systems in cities. Road networks affect traffic patterns and public transportation impacts how people move through cities and therefore where businesses spring up.

Electricity networks, gas distribution, water and sewage lines crisscross urban maps and affect what neighbourhoods might lose power first during outages or which areas are prone to flooding. And flooding can affect power outages, which then can affect public transportation’s reliability, and so on. A very comprehensive view of urban activities is required to visualise interdependencies.

One of the general hurdles to effective and efficient city management are the silos in which urban networks and services operate. Data cannot easily connect; platform and format issues prevent seamless interfacing between systems, thereby posing genuine hurdles to all-encompassing digital twins that can truly capture and reflect the operational, commercial and social ongoings within cities. Therefore, a first step to creating genuinely beneficial urban digital twins often is a rather mundane, administrative step. Collaboration between administrations and agencies is key and the need for compatible data is crucial.

The city of San Diego’s managers realised the importance of such collaboration and created a partnership between the San Diego Association of Governments; the San Diego Regional Economic Development Corporation; San Diego State University; University of California, San Diego; and industrial partners. Interconnected dynamics and challenges in urban environments require connected, relatable data and digital twins that can represent resulting complexities – the collaboration of city administrations and network users is the first step.

Cautioning against developing communal divides

Digital twins will transform the way we plan, design, operate and maintain equipment, networks and urban environments. AI will accelerate their development, improve their performance and enhance their usability. But on the road to ubiquitous use, hurdles and issues need overcoming – some considerations generally associated with virtualisation technologies and use, others relate to AI, which currently is experiencing almost unchecked excitement and investment.

The Brookings Institution recently highlighted the emerging industrial and geographic unevenness of implementing and leveraging AI. The diffusion of AI will occur on different timelines in various industrial sectors. Spending on AI will depend on productivity and economic growth that companies and industries will expect or experience. Available investment capital and shareholder agreement will also play a role.

While it is natural that technology-related companies and finance, logistics and manufacturers firms already foresee substantial changes to their operations, agriculture, mining, personal services (including some aspects of healthcare) and many government services will see less immediate application opportunities. AI’s use for digital twins of cities will therefore initially create uneven representation in various sectors of urban planning and management.

Existing disparities between countries and regions will create geographical unevenness in the use of AI, and therefore in the adoption and diffusion of AI-empowered digital twins. Research by the Brookings Institution states: “Artificial intelligence is transforming the US economy, yet regional disparities in talent development, research capacity and enterprise adoption are stark and not yet fully understood.”

The digital divide emerged as a major concern at the end of the 1990s. Although the effects did not pan out as dramatic as some observers initially warned, the Covid pandemic from 2020 and following years highlighted unevenness in the way individuals, regions and entire countries were able to move personal and commercial activities online. Geographical laggards could develop in which AI implementation is slow, leaving other regions to charge ahead.

The report continues: “Such gaps and deficits may result in unrealised opportunities for productivity growth across disparate industries, and limit discovery and dissemination of the full range of AI use cases. For that matter, disparities in AI readiness may leave some communities to fall behind or slump into ‘development traps’. Imbalances in AI talent, innovation infrastructure and business adoption very well could decide which people and places will prosper in the future – and which will not.”

Internationally, such gaps can lead to “geo-algorithmic inequality”. Digital twins that replicate commercial activities, urban environments, entire ecosystems and eventually even economies as a whole will support the development of climate-resilience strategies, affect investment flows and establish the foundation for regional development plans in developed countries and metropolises.

“By contrast, much of Africa, the Caribbean and parts of Southeast Asia remain invisible in major digital twin ecosystems,” says the report. Data availability is spotty, often non-existent. Therefore, “decisions around infrastructure aid, disaster prevention or carbon offsetting are made with incomplete information – or without them in mind at all”.

Thinking globally, acting locally

The impact on these regions can be dramatic. Geo-algorithmic inequality results in “the uneven inclusion of countries and communities in the simulations that shape global policy, investment, and resilience planning”, according to Brookings Institution.

The effect can cascade towards digital twins that attempt to simulate the global ecosystem. In such virtualisations of the entire planet, structural bias can encode misrepresentations in digital twins and therefore distort resulting applications.

Potential approaches to alleviate such concerns exist. The Gaia-X initiative is looking to establish digital sovereignty by establishing “an ecosystem, whereby data is shared and made available in a trustworthy environment”. And the World Avatar effort is working on an “ecosystem of tools and services that can be used to create an individual digital twin, or network of connected digital twins, to provide a platform of data and model interoperability”.

Data silos of networks or country initiatives can then easily connect to each other. Although laudable, a concerted policy framework is needed to create incentives for corporations and organisations to buy into and fully embrace such efforts.

Martin Schwirn is the author of Small data, big disruptions: How to spot signals of change and manage uncertainty (ISBN 9781632651921) on foresight and horizon scanning. He is a strategy and innovation consultant for Global 2000 companies.

Source

Popular large language models (LLMs) are unable to provide reliable information about key public services such as health, taxes and benefits, the Open Data Institute (ODI) has found.

Drawing on more than 22,000 LLM prompts designed to reflect the kind of questions people would ask artificial intelligence (AI)-powered chatbots, such as, “How do I apply for universal credit?”, the data raises concerns about whether chatbots can be trusted to give accurate information about government services.

The publication of the research follows the UK government’s announcement of partnerships with Meta and Anthropic at the end of January 2026 to develop AI-powered assistants for navigating public services.

“If language models are to be used safely in citizen-facing services, we need to understand where the technology can be trusted and where it cannot,” said Elena Simperl, the ODI’s director of research.

Responses from models – including Anthropic’s Claude-4.5-Haiku, Google’s Gemini-3-Flash and OpenAI’s ChatGPT-4o – were compared directly with official government sources. 

The results showed many correct answers, but also a significant variation in quality, particularly for specialised or less-common queries.

They also showed that chatbots rarely admitted when they didn’t know the answer to a question, and attempted to answer every query even when its responses were incomplete or wrong. 

Burying key facts

Chatbots also often provided lengthy responses that buried key facts or extended beyond the information available on government websites, increasing the risk of inaccuracy.

Meta’s Llama 3.1 8B stated that a court order is essential to add an ex-partner’s name to a child’s birth certificate. If followed, this advice would lead to unnecessary stress and financial cost. 

ChatGPT-OSS-20B incorrectly advised that a person caring for a child whose parents have died is only eligible for Guardian’s Allowance if they are the guardian of a child who has died. 

It also incorrectly stated that the applicant was ineligible if they received other benefits for the child. 

Simperl said that for citizens, the research highlights the importance of AI literacy, while for those designing public services, “it suggests caution in rushing towards large or expensive models, which emphasise the need for vendor lock-in, given how quickly the technology is developing. We also need more independent benchmarks, more public testing, and more research into how to make these systems produce precise and reliable answers.”

The second International AI safety report, published on 3 February, made similar findings regarding the reliability of AI-powered systems. Noting that while there have been improvements in recalling factual information since the 2025 safety report, “even leading models continue to give confident but incorrect answers at significant rates”.

Following incorrect advice

It also found highlighted users’ propensity to follow incorrect advice from automated systems generally, including chatbots, “because they overlook cues signalling errors or because they perceive the automation system as superior to their own judgement”.

The ODI’s research also challenges the idea that larger, more resource-intensive models are always a better fit for the public sector, with smaller models delivering comparable results at a lower cost than large, closed-source models such as ChatGPT in many cases.

Simperl warns governments should avoid locking themselves into long-term contracts when models temporarily outperform one another on price or benchmarks.

Commenting on the ODI’s research during a launch event, Andrew Dudfield, head of AI at Full Fact, highlighted that because the government’s position is pro-innovation, regulation is currently framed around principles rather than detailed rules.

“The UK may be adopting AI faster than it is learning how to use it, particularly when it comes to accountability,” he said.

Trustworthiness 

Dudfield noted that what makes this work compelling is that it focuses on real user needs, but that trustworthiness needs to be evaluated from the perspective of the person relying on the information, not from the perspective of demonstrating technical capability.

“The real risk is not only hallucination, but the extent to which people trust plausible-sounding responses,” she said.

Asked at the same event if the government should be building its own systems or relying on commercial tools, Richard Pope, researcher at the Bennett School of Public Policy, said the government needs “to be cautious about dependency and sovereignty”.

“AI projects should start small, grow gradually and share what they are learning,” he said, adding that public sector projects should prioritise learning and openness rather than rapid expansion.

Simperl highlighted that AI creates the potential to tailor information for different languages or levels of understanding, but that those opportunities “need to be shaped rather than left to develop without guidance”.

With new AI models launching every week, a January 2026 Gartner study found that the increasingly large volume of unverified and low-quality data generated by AI systems was a clear and present threat to the reliability of LLMs.

Large language models are trained on scraped data from the web, books, research papers and code repositories. While many of these sources already contain AI-generated data, at the current rate of expansion, they may all be populated with it. 

Highlighting how future LLMs will be trained more and more with outputs from current ones as the volume of AI-generated data grows, Gartner said there is a risk of models collapsing entirely under the accumulated weight of their own hallucinations and inaccurate realities. 

Managing vice-president Wan Fui Chan said that organisations could no longer implicitly trust data, or assume it was even generated by a human.

Chan added that as AI-generated data becomes more prevalent, regulatory requirements for verifying “AI-free” data will intensify in many regions.

Source

José Adorno/BGR

Yet another report suggests satellite 5G support might be a key upgrade for upcoming iPhone models. This time, Weibo leaker Fixed Focus Digital, who is known for revealing the iPhone 16e brand before its release, says iPhone models with the C2 modem will support 5G New Radio for Non-Terrestrial Networks (NR-NTN). With that, an iPhone would be able to connect straight to a satellite instead of just traditional ground cell towers, to take advantage of 5G capabilities.

While much still needs to be discussed, as Apple is not the only one working to make that technology mainstream, Fixed Focus Digital says this might be “a pivotal starting point for mobile phone manufacturers to enable satellite internet connectivity.”

With the iPhone 14, Apple started offering satellite features, which now include SOS Emergency Calls and Texts, the ability to be found through Find My when there’s no Wi-Fi or 5G signal, and roadside assistance. So far, this service has been free, even though the company said it would only be offered for two years without cost. However, as we move to the middle of the iPhone 17 lifecycle (more than three years since the iPhone 14 introduction), the company continues to offer everything for free and is still slowly expanding the service to more regions.

A variety of new iPhones could get this feature

José Adorno/BGR

If the rumor turns out to be accurate, Apple could bring 5G satellite connectivity to all iPhone 18 models, iPhone Air 2, and the upcoming iPhone Fold. At this moment, Apple works with Globalstar, a Starlink rival, to provide its satellite functionalities. Still, a report from The Information suggests Apple might join forces with Elon Musk’s SpaceX as Starlink uses the same radio technology that Apple has on its iPhones.

Even though a recent Reuters report suggested SpaceX might release its own phone with 5G satellite capabilities, it’s more likely that Musk might find in Apple a potential partner to promote this new era of connectivity. Then, if Apple indeed goes down that path, it would make sense if the company charged a subscription for the service, as more than paywalling a feature that could save someone’s life, it would bring 5G connectivity even in places where cellular signal doesn’t reach.

Besides that, the iPhone 18 generation is expected to get mild improvements, as no design changes have been rumored so far, and Apple will start testing the grounds with its new foldable phone. With the regular iPhone 18 and iPhone Air 2 expected for early 2027, the company is reportedly focusing on a “Pro” September launch, and nothing screams Pro like a feature no one else in the market has yet.

Source

Europe’s data protection supervisors have warned that proposals by the European Commission to reform privacy law by narrowing the definition of personal data could erode privacy rights for European Union (EU) citizens.

The regulators said in a joint response with the European Data Protection Board (EDPB) that the proposed changes raise “significant concerns” and could adversely affect the level of protection for individuals’ personal data.

The warning comes as the European Commission presses ahead with proposals to reform a raft of EU data protection laws through a “Digital Omnibus” regulation which it says will simplify compliance for businesses and boost EU competitiveness.

The EDPB and national European Data Protection Supervisors warned in a joint opinion that some of the proposed measures could damage privacy rights of individuals, create legal uncertainty and make data protection law more difficult to apply.

The contested proposals include changes to the definition of personal data that would weaken privacy rights by allowing organisations to treat personal data as non-personal data if they processed it in a way that did not identify individuals.

Proposals ‘go beyond’ European law

Although the proposals had been welcomed by many data protection practitioners as a way to simplify compliance with data protection and privacy regulations, the regulators have sounded a warning bell.

They “strongly urge” legislators not to adopt the proposed changes to personal data, arguing that they “go far beyond a targeted or technical amendment” and far beyond EU case law by “significantly narrowing the concept of personal data”.

The regulators also raise concerns about proposals that could water down individuals’ rights not to be subject to automatic decision-making by artificial intelligence (AI) or software through a proposed “exhaustive list” of cases where automatic decision-making would be allowed.

Another proposal, which would allow the European Commission new powers to determine whether pseudonymised data should no longer be classed as personal data, has also sparked calls for clarification.

The regulators warn that proposals to restrict the right of people to make subject access requests to people motivated by “data protection” concerns is not compatible with EU law.

If implemented, this proposal is likely to exclude access requests made by journalists, academics or policymakers, for non-data protection purposes, such as journalistic or academic research.

They also call for the commission to fine-tune proposals that would allow organisations to use special categories of data – including data on political opinions, religious beliefs, trade union membership, health and sexual orientation – when they are used in an “incidental” and “residual” way to train AI systems.              

Reporting data breaches simplified

The EDPB and the data protection supervisors support many of the EU’s proposals, including plans to make reporting data breaches less painful for companies.

The European Commission proposes raising the threshold of risk before companies need to make a notification and extending the deadline to file a notification from 72 to 96 hours.

“This change is not expected to substantially affect the level of protection for data subjects but would significantly reduce the administrative burden for controllers, given that they would only have to notify data breaches that are likely to result in a high risk to the rights and freedoms of data subjects,” they said.

Another proposal to offer alternative ways for people to consent to cookies to avoid “consent fatigue” and a “proliferation of cookie banners” – for example, by consenting to cookies once on a particular computer – has also been welcomed.

However, the regulators remain concerned about the proposed changes to the definition of personal data.

European Data Protection supervisor Wojciech Wiewiórowski said: “These changes are not in line with the court’s case law and would significantly narrow the concept of personal data.”

Anu Talus, chair of the European Data Protection Board, said any changes to EU Data protection law must bring legal certainty while maintaining a high level of protection of individual rights and freedoms.

“We strongly urge the co-legislators not to adopt the proposed changes to the definition of personal data,” she added. “These changes are not in line with the court’s case law and would significantly narrow the concept of personal data.”

Isabelle Roccia, managing director for Europe for IAPP, a professional association with 90,000 members, said that privacy and data protection professionals were in favour of the EU’s proposals.

“The commission proposal to narrow the scope of personal data definition was welcomed by many practitioners as a sign of pragmatism in the interpretation of the GDPR,” she said. “If adopted, it would have consequential impact in easing many friction points across contractual obligations and data transfer rules among others.

“With this joint opinion, EDPS and EDPB are signalling that they want to preserve the conservative and data-subject-first approach they have established in the past decade,” added Roccia.

She said that business leaders would also welcome legal certainty around the legal basis for when developers can use “legitimate interest” to process personal data to train AI models.

Commission proposals benefit US big tech 

Campaign group Noyb said that the “Digital Omnibus” proposed sweeping changes to the GDPR and the ePrivacy Directive that were disguised as simplification measures.

The group claims the changes would not help EU businesses that have to complete “useless” paperwork to comply with data protection laws, but would mainly be useful to big US tech companies.

Max Schrems, privacy lawyer and honorary chair of Noyb, said “the independent authorities have called out key changes for what they are: neither ‘technical change’ nor ‘simplification’, but limitations of the right to data protection for EU residents”.

Source

Having led the UK’s leading broadband provider through a decade of “unprecedented investment and transformation”, Clive Selley is to step down as chief executive officer of Openreach at the end of March 2026, to be replaced by current deputy CEO Katie Milligan.

A wholly owned and independent subsidiary of the BT Group, Openreach said that its infrastructure is already becoming the backbone of Britain’s digital economy, supporting everything from smart farming and sustainable transport to remote working and virtual healthcare.

Employing around 27,000 people, it is aiming to make its full-fibre network available to as many as 30 million premises in all corners of the UK by the end of the decade and has invested £15bn to build a new infrastructure to 25 million homes and businesses by the end of 2026. For the year up to the end of March 2025, Openreach reported revenues of £6.157bn.

Openreach claims to be building its new network to connect customers faster and further than any other provider in the UK, reaching an average of 85,000 new premises every week. In total, more than 3,500 UK towns, cities, boroughs, villages and hamlets have so far been included in the build programme, with others being reached through publicly funded partnerships.

The operator describes its build-out project as one of the largest and fastest broadband infrastructure programmes in Europe, with engineers claimed to be now reaching more than a million new homes every three months. It has so far included around 33,000 medical facilities and more than 25,000 colleges, schools and universities, helping to “transform” access to critical services.

The firm’s wholesale broadband network – the UK’s largest – supports more than 680 service providers including BT, SKY, TalkTalk, Vodafone and Zen to provide broadband, TV, phone, data and mobile services to their customers.

Milligan has been with Openreach since 2009 and, before becoming deputy CEO, was the company’s chief commercial officer, driving its commercial strategy and leading customer relationships. She is said to have played a pivotal role in accelerating the adoption of full fibre broadband nationwide, as well as strengthening relationships with Openreach’s partners, customers and stakeholders.

Commenting on his decision to step down, Selley said that it has been an honour to deliver on a mission to connect the UK building what he said was a new digital platform for growth and prosperity across the country. “Openreach is a remarkable business with amazing, talented people – and I’m incredibly proud of what we’ve achieved together,” he said.  

“I’m also delighted that Katie will now be taking the reins. She knows our business inside-out, she has the right set of skills and experience for this role, and she’s passionate about delivering for our people, our customers and the nation. There is nobody I’d trust more to take care of the future and Openreach has a strong future under her leadership.”

BT Group chief executive Allison Kirkby added: “Openreach is a critical national asset – the digital backbone of the UK – and a key driver of BT Group’s long-term value. Clive’s contribution at the helm of Openreach has been exceptional. His leadership – particularly the scale, pace and quality of the full fibre broadband build – has set new standards for our industry.”

“We are deeply grateful for the commitment, expertise and integrity he has brought to the role. Clive’s lasting legacy is a world-class digital infrastructure that will serve the UK for generations to come.”

Source