Getty Images/iStockphoto/monsitj Balancing strained budgets with endpoint modernization demands Commentary Read more NIST SP 800-223 and 800-234: A turning point for federal high-performance computing security Commentary Read more Getty Images/Phimprapha Kitaiamphaisan From mandate to momentum: Turning CISA’s edge device directive into lasting capability Commentary Read more
News EVT advances smart manufacturing with AI-powered Machine Vision Solutions Eye Vision Technology (EVT), an international specialist in industrial machine vision and image processing software, is helping manufacturers achieve higher levels of quality, efficiency, and automation through its advanced EyeVision platform. With more than 25 years of expertise in machine vision, EVT has established itself as a trusted partner for companies seeking to implement reliable, real-time inspection systems. Its solutions are designed to support the growing demand for zero-defect production while reducing engineering complexity and operational costs. At the heart of EVT’s offering is the EyeVision software platform, a powerful and flexible system that enables users to configure sophisticated image processing applications without the need for traditional programming. Featuring an intuitive drag-and-drop interface, the software allows engineers and production teams to quickly develop, test, and deploy inspection routines, significantly shortening implementation times. Combining AI and Traditional Image Processing One of the key strengths of EyeVision is its ability to integrate both conventional rule-based image processing and modern artificial intelligence techniques within a single environment. This hybrid approach allows users to choose the most appropriate method for each application, ensuring optimal performance across a wide range of inspection tasks. The platform’s AI and deep learning capabilities are particularly effective in detecting complex or variable defects that would be difficult to identify using traditional methods alone. At the same time, established image processing tools provide fast and reliable results for structured and repeatable inspection processes. Core capabilities of the EyeVision platform include: - AI-based defect detection and classification - High-precision 2D and 3D measurement and metrology - Vision-guided robotics for automated handling and assembly - Multispectral and thermal imaging for advanced analysis - Optical character recognition (OCR), barcode, and code reading for traceability Together, these features enable manufacturers to maintain consistent quality
Dutourdumonde - stock.adobe.com UK High Court dismisses facial-recognition judicial review case The Metropolitan Police has won a judicial review case that argued its live facial-recognition policy was unlawful The UK High Court has dismissed a judicial review case against the Metropolitan Police’s use of live facial-recognition (LFR) technology, ruling there are sufficient constraints in place to prevent abuse and ensure compliance with human rights law. Brought by anti-knife campaigner Shaun Thompson, who was wrongfully identified by the Met’s system and subject to a prolonged stop as a result; and Silkie Carlo, the director of privacy group Big Brother Watch, the landmark legal challenge argued there are no meaningful constraints on how the Met can deploy the technology. In particular, their challenge hinged on the argument that the Met’s policy on where it can be deployed and who it can be used to target is so permissive, and leaves so much discretion to the force, that it cannot be considered “in accordance with law”. However, the High Court ultimately agreed with the Met’s lawyers that “the Policy contains adequate and lawful constraints” over how and where the technology can be used. While the Court of Appeal ruled in August 2020 that the use of LFR by South Wales Police was unlawful because the policy in place left excessive discretion in the hands of the force, the High Court found that in the Met’s case, its policy contained clear deployment criteria that effectively prevents individual officers from acting on “whim, caprice, malice or predilection”. Although Thompson and Carlo argued that the Met’s policy could lead to disproportionate deployment rates in areas with large ethnic minority communities, the court said it “heard no developed or meaningful challenge on discrimination grounds” that would allow it to accept this argument. It added although a properly
Legal safeguards vital if facial recognition tech is rolled out by Police Scotland A year has passed since the conclusion of what Police Scotland described as a “national conversation” about its potential use of live facial recognition (LFR), but it remains far from clear if, let alone when, the controversial technology will be adopted. The artificial intelligence-powered tech, which allows scans of human faces to be compared against police databases, is hailed by some justice figures as a crucial biometric tool in the fight against crime, and one that can be used in a justifiable and proportionate manner. Advertisement Hide AdAdvertisement Hide AdBut growing ethical concerns around algorithmic bias and accuracy have prompted a backlash, with human rights campaigners warning of a “historic shift” in the relationship between Police Scotland and the public. Not as effective as claimed Across England and Wales, LFR is being used by 13 forces, with the Met significantly scaling up its deployments; in the first four months of 2026, it has scanned more than 1.7 million faces. As far as key watchdogs are concerned, the rapid increase in LFR’s use presents significant problems. William Webster, the biometrics commissioner for England and Wales, cautioned this week that the “slow pace of legislation was trying to catch up with the real world” while Dr Brian Plastow, the Scottish biometrics commissioner, said the technology was “nowhere near as effective as the police claim it is”. Advertisement Hide AdAdvertisement Hide AdAlasdair Hay, chair of the Scottish Police Authority’s policing performance committee, has stressed that while no decision has been made, Scotland’s national force would be required to produce a bespoke code of practice, and demonstrate compliance with the biometrics commissioner’s own code. Police Scotland’s conversation exercise, meanwhile, found the public was split. While 49 per cent of respondents to
Meta uses AI profiling to infer user age, enforce teen restrictions Meta says it has begun using AI to detect and remove users under 13 from its platforms, and to automatically turn on Teen Account protections for users it deems to be teenagers. A blog post says the algorithmic function will flag Instagram users in Brazil and the EU, and Facebook users in the U.S., even if they have declared an age over 18. The firm also wants you to know it is “providing parents on Facebook and Instagram with tips on how to help them have conversations with their teens about the importance of being honest about their age online.” Meta has put on a good show of caring about online safety and age assurance legislation. The company is continually explaining how it supports the goal of keeping people safe online; its latest missive claims that, “for over a decade, we’ve built tools, features, and resources to help teens have safe, age-appropriate experiences on our apps. This includes launching Teen Accounts on Instagram, Facebook, and Messenger with built-in protections that limit who can contact teens and the content they see.” “For years, we’ve worked to find and remove accounts that belong to those we believe are underage,” Meta says. “Today, we’re providing more detail on our ongoing efforts to develop advanced AI that detects underage accounts, including the use of visual analysis to look beyond simple admissions of age.” Australia has proof Meta is bad at keeping kids off its platforms The suggestion that Meta works hard to keep kids off its services is easily exploded with reference to two bits of context: first, data from Australia showing that Meta has done a miserable job of complying with its Social Media Minimum Age, and second, court decisions in the
Meta is unleashing AI that scans users’ bodies — from face shape to height — in an aggressive bid to root out underage accounts on Facebook and Instagram. The company announced Tuesday it was developing “advanced AI” that includes the use of visual analysis for detecting underage accounts. This new visual analysis technique will enable Meta’s AI to scan photos and videos for “visual clues” about a user’s age – including one’s height and bone structure. These visual insights will be combined with Meta’s analysis of text and interactions in a bid to significantly increase the number of underage accounts it can spot and remove. Meta said it was also adopting AI-powered textual analysis to analyse entire profiles for contextual clues – such as birthday celebrations or mentions of school grades – across posts, comments, profile bios, captions, and in-app features such as Instagram Reels, Instagram Live, and Facebook Groups. “If we determine an account may be underage, it will be deactivated and the account holder will need to verify their age to continue using their account,” a Meta spokesperson said. Australian rollout coming Following a testing period in the US, Meta has planned to globally expand its visual analysis technology to under-13s over the coming months. A spokesperson confirmed to Information Age that an Australian rollout for under-16s will follow, bolstering Meta’s compliance with the country’s social media minimum age (SMMA) laws. “In Australia, we are committed to complying with the social media ban and are extending these advanced detection tools to under-16s in the coming weeks and months,” said Antigone Davis, vice president and global head of safety at Meta. As of January, Meta had blocked more than 544,000 under-16s from its platforms in Australia. Further to removing accounts for underage users, Meta said it would also expand
POLICE vans using facial recognition software were out and about in Winchester. Officers joined forces with the specialist Live Facial Recognition (LFR) team on Tuesday, May 5, on the deployment of two LFR vans in the High Street. This technology is being used by the police to help officers identify suspects for high-risk and priority offences, as well as help find any missing people. Read More - Police spotted in Winchester looking for areas to use facial recognition vans - Architecture firm shortlisted for four categories at property awards A police spokesperson said: "This sort of precision policing that sees the latest technology working in the background, targeting those most wanted criminals, allows our officers to focus more on responding to emergencies, spending time patrolling and investigating crime that matters most to the local community. "No arrests were made as a result of this deployment." More information about how Live Facial Recognition works can be found on the Hampshire Constabulary website.
KERV.ai has bet that the future of advertising won’t be defined by ad placements—but by the moments that actually matter. The company has introduced Moment Match Engine, an AI-powered solution that identifies the most meaningful moments in video content and aligns them with brand and product signals, unlocking greater monetisation value for publishers, improved engagement for advertisers, and seamless commerce and brand discovery for viewers. Rather than treating content as a singular opportunity for delivery, KERV.ai’s approach captures moments when attention peaks and intent signals emerge—turning those high-value instances into opportunities for connection. Through seamless integrations with leading publishers, retailers, and ad platforms, the Moment Match Engine improves engagement by enabling existing media to work with interactive commerce experiences in real time, driving measurable outcomes. KERV.ai’s proprietary AI and image recognition technology ingests and analyses both video-on-demand and live video content to generate actionable signals that identify high-value moments when consumers are most engaged and likely to convert. Powered by patented product recognition technology, KERV.ai enriches existing video at scale—down to the pixel level—uncovering the true meaning of each scene. This creates new layers of contextual signals that understand not just what’s in the content, but what matters within it. The result is a system that enables advertisers to appear in the moments most likely to drive action, without disrupting the viewing experience or compromising brand safety or suitability. For publishers, Moment Match Engine introduces a new model centred on diverse, scalable value creation within content itself. By leveraging deep contextual signals, publishers can surface brand and commerce experiences that feel additive to the viewing experience while maintaining full control through advanced metadata validation layers and industry-trusted compliance frameworks. “Historically, advertising has been about inserting messages into content,” said Gary Mittman, CEO of KERV.ai. “We’re shifting that model to align
- When The Omaha Oracle Calls It A Casino, You Pay Attention - The $397 Billion Question Of Berkshire Hathaway - The Valuation Problem No One Wants to Talk About - What This Means For Indian Investors At the Berkshire Hathaway annual meeting on May 2, 2026, the most watched investor alive sat in the audience for the first time in 60 years and still managed to steal the show. Warren Buffett, now chairman emeritus, looked at the current US stock market and called it what many are reluctant to say out loud: a casino. "We've never had people in a more gambling mood than now," he said. Let's break down what Buffett and Berkshire Hathaway have to say about where markets stand today, how the world's most admired conglomerate is positioned, and what it all means if you're an Indian investor watching US stocks. When The Omaha Oracle Calls It A Casino, You Pay Attention Buffett was not mincing words. When asked why Berkshire wasn't deploying capital despite the market volatility of 2026, he was direct. He specifically called out one-day options and prediction markets as behavior that has crossed the line from investing or even speculating into outright gambling. This is not just colorful language from a 95-year-old. It is a signal rooted in decades of pattern recognition. The last time Buffett described markets in terms this cautious, the conditions that followed were not kind to investors who ignored him. His benchmark for deploying capital is clear. He needs to see genuine fear, not a 10% pullback, but the kind of dislocated, panicked selling that happened in 2008 or during the Covid crash of March 2020. A volatile but fundamentally supported market does not meet that bar. The $397 Billion Question Of Berkshire Hathaway Here is the number
'No money for new weapons' and 'Cost of pint hits £10' The Guardian carries warnings about facial recognition technology, widely used by many police forces and a growing number of retailers. The biometrics commissioner for Scotland tells the paper that the technology is "nowhere near as effective as the police claim it is". Along with his counterpart in England and Wales, he is calling for new laws to govern how and when the technology is used and a new regulator to clamp down on misuse. The co-author of the strategic defence review, Gen Sir Richard Barrons, tells the Times that the armed forces will have no money for new weapons until 2030. He says there is "just about" enough funding for tanks and helicopters but not enough for unmanned or AI-assisted weaponry. However an army source disputes the claim, telling the paper money is already pouring into rapid procurement programmes. The Sun says an investigation has revealed that workers in Pakistan making the official Adidas football for this summer's World Cup are making as little as £26 a week. The priciest version of the ball sells for £130. "The Beautiful Shame" is the paper's headline. Adidas tells the Sun all its products are manufactured under fair and safe working conditions. Several papers report on the new injectable form of a cancer drug being rolled out across the NHS. A senior doctor tells the Daily Telegraph the jab will offer a "lifeline" to thousands of patients, giving them the freedom to live their lives instead of spending hours in a hospital. "This shows what happens when innovation meets determination," The Mirror's editorial says. The vast cost of financing the construction of the data centres required for artificial intelligence makes the front page of the Financial Times. The paper says that banks
We're all too familiar with the notch—the unsightly cut-in that graced many smartphones for years, like the iPhone X or the LG G7. The notch has largely been replaced on today’s smartphones by floating punch-hole cameras that take up less space and look a little more futuristic, though notches are still prevalent on some laptops, like Apple’s MacBooks. On the iPhone, Apple calls its floating pill-shaped camera system the Dynamic Island, which debuted on the iPhone 14. The iPhone still has the largest camera cutout today, due to its Face ID biometric authentication system. (Barring Google Pixel phones, the vast majority of Android phones don't offer a secure face authentication equivalent, so they don't need a bulky camera cutout.) This island could get much smaller, however, thanks to new under-display camera technology announced at Display Week 2026 from Metalenz, a optics startup from Boston. A Primer on Metasurfaces Metalenz’s optical metasurfaces technology is a flat-lens system that uses a fraction of the space of traditional multi-lens elements in most smartphones. You can read more about it in our original coverage of the company here, but in short, instead of refracting light through multiple plastic or glass lens elements—which improves image clarity, corrects aberrations, and brings more light to the camera sensor—metasurfaces use a single lens with nanostructures to bend light rays toward the sensors. Metalenz says more than 300 million of its metasurfaces are already used in consumer devices today, replacing bulky traditional optics in time-of-flight sensors that capture depth information and assist with a camera's autofocus. The company also pioneered a method to use these metasurfaces to capture polarization data. When light hits an object with specific material properties, it creates a unique polarization signature. Light reflecting off black ice has a different polarization signature from light reflecting off
When Henry VIII married Anne of Cleves, historical records state he was so appalled that her beauty did not match the portrait by Hans Holbein the Younger that his fourth wife became widely known as the “Flanders mare”. Now, Holbein’s pictures of another of Henry’s wives, Anne Boleyn, have been called into question using a more scientific method of analysis: artificial intelligence (AI). Researchers used facial recognition tools to establish that a sketch of a previously unknown woman could be of Henry VIII’s doomed second wife. Meanwhile, a drawing previously thought to be of Boleyn by Holbein in the Royal Collection has probably been misidentified and is more likely to be a picture of her mother. Academics used AI to re-examine the second drawing, known as the Windsor Sketch, which for 200 years has been assumed to be the Tudor queen based on an 18th-century inscription that says it is “Anna Bollein Queen”. The portrait, labelled as Anne Boleyn (c.1500-1536), depicts a blonde woman with a double chin. This contradicts contemporaneous accounts of Boleyn, which describe her as slender, dark-haired and with a distinctive “little neck”, a feature she reportedly referenced before her execution. Using facial recognition software, experts went on to suggest that another drawing by Holbein the Younger in the Royal Collection is more likely to be Boleyn. Karen L Davies, a historian who led the study, which is published in Nature Heritage Science, said: “I’ve spent many years researching Anne Boleyn, and what ultimately drove me to this project was a growing dissatisfaction with the long-accepted identification. There were simply too many contradictions. “The drawing didn’t align with the primary sources and each proposed explanation seemed only to generate further inconsistencies rather than resolve them.” Davies and academics from Bradford and Stanford Universities set out to use
A new report reveals that children across the UK are outwitting online safety measures with fake birthdays, borrowed IDs, and some surprisingly creative facial hair. A third of children say they have bypassed online age checks in the past two months - some by drawing fake moustaches on their faces to trick facial recognition software. The report from Internet Matters titled The Online Safety Act: Are Children Safer Online? surveyed 1,270 children aged 9-16 and their parents across the United Kingdom to see whether the country's landmark online safety legislation is delivering any meaningful protection for children. One mother told researchers she caught her son using an eyebrow pencil to draw a moustache on his face to pass a platform's facial age estimation check. It worked. He was verified as 15. He was 12. What did the report find out? The study discovered that 46% of children believe age checks are easy to bypass, while only 17% say they are difficult. Among the infiltration methods children described were entering a fake birthdate, using someone else's identification, submitting videos of other people's faces, and using video game characters to fool facial recognition tools. "I've seen clips of people online where they'll get clips of video game characters like turning their head and use it for age verification," one 11-year-old girl told researchers. Older children were more confident about circumventing checks, with 52% of those aged 13 and over saying age verification is easy to beat, compared with 41%of those aged 12 and under. The most common reasons children gave for bypassing age checks were to access a social media platform they were not old enough to use (34%), to join an online game or gaming community (30%), and to use a messaging app (29%). The report also found that just over
Trivia Quiz·Disney Quiz·Posted 15 hours agoYou Think You Know Disney? I Bet You Can’t Even Name Half Of These MoviesGuess the movie: ❄️👸🏰by Lauren GarafanoBuzzFeed StaffcommentFacebookPinterestLink Here's the deal: I turned all your favorite Disney movies into emojis. Some are suuuuuuuper easy. And some...well...good luck! Do you think your pattern recognition skills can help you correctly identify these Disney movies? Let's find out! Share This QuizFacebookPinterestLinkComments Comments
Abstract We propose ADGNET, a semi-supervised framework for Alzheimer’s disease (AD) diagnosis that jointly optimizes image reconstruction and classification through shared feature representations. The architecture integrates a residual backbone with attention modulation for dynamic feature selection, an encoder-decoder reconstruction branch for unsupervised representation learning, and a classification branch with focal loss to address class imbalance. This dual-task design enables effective feature learning from limited annotations. On two public MRI datasets—KACD (2D, 6,400 images) and ROAD (3D, 532 scans)—ADGNET achieves average performance improvements of 4.1% and 7.2% over state-of-the-art methods (ResNeXt WSL, SimCLR) across six metrics. Interpretability analysis using Grad-CAM and attention visualization confirms that the model focuses on clinically relevant neuroanatomical structures, particularly the hippocampus and temporal lobes, with strong correlation to established AD pathology (r = 0.67, p < 0.001). These results validate the model’s exceptional generalization capability and feature representation effectiveness across multi-modal medical imaging data, offering an efficient solution for few-shot medical image analysis. Citation: Yang X (2026) Reducing annotation burden in medical imaging with ADGNET: A semi-supervised deep learning strategy. PLoS One 21(5): e0348596. https://doi.org/10.1371/journal.pone.0348596 Editor: Vince Grolmusz, Eotvos Lorand University: Eotvos Lorand Tudomanyegyetem, HUNGARY Received: September 6, 2025; Accepted: April 17, 2026; Published: May 4, 2026 Copyright: © 2026 Xiaobo Yang. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited. Data Availability: All relevant data are within the manuscript and its Supporting Information files. Funding: The study is supported by Zhejiang Province Natural Science Foundation, grant number Y1110023 to XY. The funder had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. Competing interests: The authors have declared that no competing interests
If you’ve tried using Claude to learn to code and walked away feeling like you just witnessed a magic trick rather than actually learned something, you’re not alone. Without structure, AI coding tools default to doing the work for you—you end up with hundreds of lines of code without understanding the algorithm or the logic behind it. That said, after much trial and error, I’ve finally perfected a prompt that turns Claude into an exercise-driven tutor. It focuses on teaching concepts by making you work through the code, while also remembering your progress and difficulties to refine future sessions. Best of all, it only takes two minutes to set up. How you can use Claude vibe coding to learn actual coding The daily workflow and context tracking It’s not exactly news that you can use an LLM as a tutor. Being able to ask questions in natural language and get a relevant, contextual answer is a genuine game changer for self-learners. The experience is fundamentally different from—and often better than—Googling something. With search, you had to frame everything as keywords and hope someone had already written about your specific problem in a way that made sense. With an LLM, you just ask, and it explains the concept in a way you can actually relate to. Artificial intelligence basics Trivia challenge From chatbots to neural networks — find out how much you really know about AI. What does the term 'machine learning' most accurately describe? Who is widely credited with coining the term 'artificial intelligence' in 1956? What type of AI model powers popular chatbots like ChatGPT? What is 'overfitting' in machine learning? What is 'AI bias' most commonly referring to? What does 'GPT' stand for in AI model names like GPT-4? Which of the following best describes 'deep learning'? What
by Matthew R. Jewell When many hear the term “AI,” their thoughts go to some massive computer system that has all the answers. In reality, AI is an umbrella term that includes a variety of computer/software systems. Today, the most notable are Large Language Models (LLMs), such as ChatGPT. Not as well known, but in high use, are Robotic Process Automations (RPAs). While there are other AI tools performing facial recognition or producing songs, LLMs and RPAs have the potential for the most applicability to contracting and, therefore, adding to the contracting toolbox. An LLM can assist contracting professionals in producing the many documents that make up a contract file. Through a series of questions and responses, or prompts, a properly trained or configured LLM can assist the contracting professional in writing just about any document. With regard to RPAs, we already have an example in the contracting toolbox with the Determination of Responsibility Assistant (DORA) bot. Simply put, RPAs automate an existing process. Much like other tools before them, AI tools reduce the toil and time associated with completing a task. In this regard, they are no different than Microsoft Excel or computer aided design (CAD). Not specific to a particular industry or profession, both assist those with knowledge, skill and ability in a particular field to complete their work more efficiently, with increased reliability and quality. These tools do not replace the user’s core competencies, rather, they enhance their abilities. AI tools can generate something “new” or identify patterns that are not readily discernable to a human. Imminently useful across domains, these aspects are especially helpful to the contracting professional. With a properly trained or configured LLM and good prompt engineering, a contracting professional can produce a new product in a fraction of the current time required. Note
The Troy City Council has drafted a local law establishing standards for the use of automated license-plate reading cameras following its recent conflict with Mayor Carmella Mantello over the city's contract with Flock cameras. Crafted by Councilperson Noreen McKee, the local law will be introduced for discussion at the council's Thursday meeting. A public hearing on the legislation is tentatively set for 5:30 p.m., May 21 at 5:30 in Troy City Hall. Mantello, the Rensselaer County district attorney and Troy police chief came out against the measure on Monday. The council says the camera systems collect detailed data on people without their knowledge and poses risks to privacy, civil liberties and the freedom of movement if unregulated. To "protect residents from government surveillance and to maintain public trust in city operations," the proposed law would limit the city's use of automated license-plate reading (ALPR) cameras and the transfering of plate data to certain conditions, such as in connection with an investigation, a crime or a missing person. It would require plate data obtained by the city to be permanently deleted within 48 hours, except in certain legal situations. It would also require any city department or agency that operates or uses the system to post an annual report on the city's website and maintain a log of every query for three years. “The council supports the important work of the Troy Police Department to investigate crime in our community, but as this technology continues to rapidly evolve, we must ensure protections and oversight are in place,” council president Sue Steele said. The legislation also sets a penalty for persons injured by violations of the proposed law by entitling them to damages for mental pain and suffering, or $1,000, whichever is greater, reasonable attorneys' fees and costs of litigation. The law
The Syracuse Common Council will vote on its biometric law in two weeks, after it was tabled at their meeting on Monday. That law would prevent businesses from using biometric identification surveillance systems. According to the American Civil Liberties Union (ACLU), those systems can collect and analyze various information, including the shape of people's faces and eyes, along with their voice and even how they walk. "Biometric scanning makes a lot of mistakes based on people that look a little bit alike, but imagine were it makes mistakes where people don't look anything alike, and they're just tracking skin color or how far apart your eyes, the size of your lips, the size of your nose, it's scary to me," Jimmy Monto, 5th District councilor, said. Banks would be exempt from the law given safety concerns. The addendum to existing laws was sponsored by Councilors Monto, Chol Majok and Corey J. Williams. Erie County passed a similar law last week. The proposal comes after Wegmans recently deployed cameras equipped with facial recognition technology at select stores, saying some stores are using the technology to help with misconduct and retail theft, and that it doesn't collect other biometric data and disposes of the images and video after security purposes are fulfilled.
It happened in a flash outside Barclays in Croydon town centre. A digital trap snapped shut around one of Britain’s thousands of wanted criminals. In little over a minute, a combination of high-definition cameras, automated AI face scanning and half a dozen police officers had run a wanted man to ground. After the handcuffs clicked shut, the Metropolitan police’s controversial live facial recognition (LFR) cameras had chalked up another arrest: the fifth in 45 minutes on a regular Thursday morning. The arrest was one of hundreds made during a six-month Met police pilot of LFR cameras on vans and fixed to lamp-posts, as also seen across cities in China, the UAE, India and Israel. Critics have called the technology invasive, unregulated and anti-democratic, cited studies suggesting racial bias and called for it to be scrapped. But the Met police commissioner, Mark Rowley, has said it is “gamechanging” and keeps the public safe. The trap was set at 10am, when the clusters of surveillance cameras mounted high on pillars at the junction of Church Street and North End were switched on. Standing nearby was Kevin Brown, a plain-clothes police sergeant. The wanted man unwittingly walked past one of the cameras and instantaneously Sgt Brown’s handheld beeped furiously and flashed up the suspect’s pin-sharp live photo, alongside a previous custody photo, his name, the suspected crime and warnings of any weapons or drugs risk. Every face passing the cameras – as many as 5,000 an hour – was being scanned and its biometric data streamed live to a police operations room five miles away in Sydenham. There, an AI-powered system, supplied by the Japanese tech company NEC, checked it instantly against photos of wanted suspects and people under court orders. Brown had a match. Uniformed officers standing across the road received the