Lessons from building reliable LLM inference infrastructure by Ying Chen, Wendy Hu, Ankit Mathur, Mike Eastham, Pei-Lun Liao, Wai Wu and Arjun DCunha At Databricks, we’ve built a unique inference platform that serves every frontier model, from open source models like Kimi and Qwen to proprietary models like OpenAI, Gemini, and Claude. We power inference for some of the largest agentic applications in the world, including Superhuman, Yipit Data, Fox Sports, and others. Today, we serve more than 120T tokens per month. What makes LLM serving hard at scale is reliability. With agents becoming the interface to how we work and live, inference demand is growing exponentially. We see extremely spiky demand curves that peak during working hours. What does it mean to be a reliable inference platform? The contract appears simple. Availability is whether the request can be processed. But, in practice, different use cases have significantly different latency requirements, and this factors into availability. The most advanced agents cannot afford for p95 time to first token (TTFT) and output tokens per second (OPTS) to degrade. In a multi-tenant system for LLM serving, achieving both reliability and latency is challenging. Reliability Frontier performance requires the latest GPUs with high bandwidth interconnect for KV cache transfer. These compute setups are fundamentally less reliable than classical CPU systems, and they are expensive. Given that all-to-all communication is required,, a single node’s downtime requires reconfiguration for multiple other nodes in disaggregated prefill/decode setups. The highest bandwidth networking requires single-spine connectivity in a single physical rack (e.g. NVL72 systems). This means failures in specific systems within a single datacenter rack can create a wide-blast-radius outage. Standard tricks in distributed systems like multi-AZ or leveraging backup instance types mean keeping expensive backup GPUs idling, a cost-prohibitive option. Overprovisioning is another classic trick, but given
May 27, 2026 · via databricks.com
CasaPerks Technologies, Inc. Raises $15.8M Seed CasaPerks Technologies raises a $15.8M seed round led by Longevity Equity to scale its AI-powered resident loyalty and workplace recognition software platforms. CasaPerks Technologies, Inc. Raises $15.8M Seed CasaPerks Technologies, Inc., an Austin, TX-based AI-powered intelligent rewards platform, has closed $15.8 million in seed round. Investors The funding round was led by Longevity Equity, a private investment vehicle headed by Will Steakley. Additional participation came from institutional investors and various top real estate operators. CasaPerks Technologies Use of Funds Capital from this seed round will be used to accelerate the company’s expansion across student housing and conventional multifamily real estate sectors. The funds will also scale the WorksPerks AI-powered workplace recognition platform, invest in the company’s sales engine, and grow its brand partnership ecosystem to further develop its consumer rewards business. About CasaPerks Technologies Led by CEO Kevin J. Bradt, CasaPerks Technologies provides an AI-powered rewards platform designed to help organizations improve resident and employee engagement. The company develops software solutions, including CasaPerks for resident loyalty in student housing and multifamily properties, and WorksPerks for employee recognition in small and medium businesses, both built on Anthropic’s Claude. By leveraging AI to personalize rewards and surface retention insights, the company aims to help property managers and business owners increase activation, improve retention, and reduce turnover. Funding Details Company: CasaPerks Technologies, Inc. Raised: $15.8M Round: Seed Funding Date: May 26, 2026 Lead Investor: Longevity Equity Additional Investors: Institutional investors, top real estate operators Company Website: https://www.casaperks.com/ Software Category: SaaS Source: https://www.morningstar.com/news/business-wire/20260526724861/casaperks-technologies-inc-ai-powered-rent-and-workplace-rewards-platform-closes-significant-158m-seed-round
May 27, 2026 · via thesaasnews.com
Artificial intelligence | UK Regulatory Outlook May 2026 Published on 27th May 2026 UK updates: King's Speech 2026: AI aspects | Crime and Policing Act 2026: AI-related provisions | ICO sets out five steps to combat AI-powered cyber threats | Government publishes response to AI and copyright report | EU updates: EU legislators reach provisional agreement on Digital Omnibus on AI | Commission consults on draft guidelines for the classification of high-risk AI systems under the EU AI Act | Commission opens consultation on draft guidelines on AI transparency obligations under the EU AI Act UK updates King's Speech 2026: AI aspects King Charles III opened Parliament on 13 May 2026 with the announcement of 37 bills his ministers would like to pass in this parliamentary session. While there is no mention of any plans for further regulation of AI (including in respect of the continuing debate around AI and copyright), the new Regulating for Growth Bill will put regulatory "sandboxes" onto a statutory footing to allow businesses across the economy, through the temporary relaxation of existing rules, to test innovative AI products and other emerging technologies safely, in a real-world setting. The Police Reform Bill will deliver reforms to policing by, among other things, "equipping the police with the technology and skills [they need]". This includes the introduction of a new legal framework governing the use of facial recognition and similar technologies. The framework will set out the situations in which the use of these technologies is justified and create a single, expert independent regulatory body to provide independent oversight and advice. See this Insight for more on these and other announcements in the King's Speech. Crime and Policing Act 2026: AI-related provisions The Crime and Policing Bill received Royal Assent on 29 April, becoming the Crime and Policing
May 27, 2026 · via osborneclarke.com
5 Ways To Tell If Someone's Recording You With Smart Glasses As cool and convenient as smart glasses may seem conceptually, their actual use in the real world comes with a very real concern: People recording your private moments with smart glasses. In a time when it's becoming harder to keep a lid on your digital privacy, having your real-life privacy violated in this manner can be extremely frustrating, especially as brands like Meta upgrades its smart glasses with features like facial recognition. While there are equally high-tech solutions to this problem, like smart eyewear detection apps, if you don't have your phone handy, the only other option is to be vigilant for signs of recording, like obvious camera lenses, suspicious motions, and audio cues. While smart glasses are generally designed to be discreet, many models include tells in their construction, some subtle, some overt. Nice as it would be to not think about something like this, you have to consider your privacy in the age of smart tech-powered surveillance, and that means keeping a sharp eye out for the warning signs that someone is capturing your image without consent. Spot a recording light In the best-case scenario, whatever brand of smart glasses an onlooker is wearing included privacy warning features into its design. In the same way that a traditional handheld camera has a red light indicating that it's recording, so to do many models of smart glasses. Major manufacturers like Ray-Bans and Oakley place small, yet distinct indicator LEDs on the front of their frames, usually in one of the corners. Ideally, this LED will be clearly illuminated whenever the glasses' built-in camera is actively capturing stills or footage. Unfortunately, this helpful warning light is not a given. Not only is it not present on all models of
May 27, 2026 · via bgr.com
The Swiss venture capital community SwissVC, founded in 2015, is repositioning itself to a smaller, structured network for full-time venture capital investors deploying capital in Switzerland. The change comes with new leadership, a board overhaul and a plan to incorporate as a non-profit association. Over 11 years, SwissVC grew into a network of more than 500 venture capital professionals across Switzerland, Europe and the US. "The old model created energy, but not enough operating leverage," said Raph Grieco, incoming lead of SwissVC. "What active investors need now is not more ambient ecosystem activity. They need sharper peer exchange, better pattern recognition, and trusted rooms where real operating questions get worked through." The repositioning centers on a system of working groups, which SwissVC calls guilds, around three areas: deal intelligence, firm architecture, and portfolio management and returns. Each guild is intended to bring together eight to 15 members through a mix of invitations and applications. Legal, tax, fund-operations and AI specialists will be brought in to work directly with the guilds. The new SwissVC is aimed first and foremost at active, deploying, full-time venture investors in Switzerland. This means the "people" in the new SwissVC are not a general audience; they are practitioners with specific operating questions, relevant pattern recognition, and a willingness to contribute to shared outputs. The model explicitly rejects passive membership inside the guild structure and treats contribution as a requirement for access to the highest-value rooms. SwissVC also plans to scale back its programming to one flagship event per year, alongside a calendar of working sessions, and to split membership options into different tiers depending on adherents’ level of contribution to the ecosystem. The community will be led by Raph Grieco, who launched UPCOMINGVC, an educational platform for aspiring venture capitalists, in 2017 and then went on
May 27, 2026 · via startupticker.ch
You do not have permission to retrieve the URL or link you requested,
If you think this was a mistake please call 1300 134 174 or e-mail customercare@realestate.com.au
and quote the reference number #18.aa4c017.1779961346.8abf4bf0
May 27, 2026 · via realestate.com.au
17% of initial adult decisions later found to be children, with hundreds of cases still pending To accompany new data on age assessments in last week's immigration statistics, the Home Office published a one-off report looking at how many asylum seekers have their age assessed. Image credit: UK GovernmentThe report, which is available here, shows that a significant proportion of asylum seekers initially classified as adults by immigration officials are later determined to be children. The Helen Bamber Foundation notes it is the first time the Home Office has published detailed outcomes on age disputes. Figures in the Home Office report cover the period from July to December 2025. They show that border officials made Initial Age Decisions (IADs) that 1,885 individuals claiming to be minors were adults. Of those, 949 subsequently underwent a full age assessment by a local authority or the National Age Assessment Board (NAAB). To date, 326 of these individuals (17% of the original 1,885) have been confirmed to be children, while 377 cases remain pending. The Home Office noted that the number of people undergoing age assessments has risen sharply in recent years, reaching 6,420 individuals in the year ending March 2026. This means 7% of all asylum claimants were subject to an age assessment, compared to between 1% and 3% annually before 2020. The report attributes this increase to the rise in arrivals via small boats, who are less likely to hold documentation proving their identity. The statistics show a distinct variation in outcomes depending on the assessing body. Between July 2025 and March 2026, 51% of IADs conducted by border officials resulted in an adult classification. In contrast, local authority assessments determined the individual to be an adult in only 28% of cases, finding them to be a child in 68% of instances.
May 27, 2026 · via ein.org.uk
Facial recognition vans will be deployed in Bicester this Friday.
Thames Valley Police will deploy Live Facial Recognition vans in Sheep Street on Friday, May 29.
The specialist unit will work alongside local officers to identify known suspects, deter crime and help keep the community safe.
READ MORE: Drugs still being dropped by drones into UK prison
The force says informing the community in advance is part of a commitment to transparency, even though it may lead some individuals to avoid the area.
A live CCTV feed captures facial images before they are input into the system.
The Live Facial Recognition technology then analyses facial features in real time to create a biometric template against a predetermined watch list of people suspected of crimes or wanted by the courts.
May 27, 2026 · via oxfordmail.co.uk
The Department of Homeland Security is expanding its capacity to scan irises as part of its mass deportation efforts, a move that has raised concerns among privacy experts that the agency, flush with an influx of funding, is gathering biometric data from people it detains. The agency awarded a $25 million no-bid contract last week to BI2 Technologies, a company that specializes in iris scanning. The new contract is more than five times the amount of the company's last DHS contract, awarded last fall. NPR reached out to BI2 multiple times regarding its work with ICE, but did not hear back. As part of its proposal to the company, DHS requested more than 1,500 iris scanners, as well as access to the company's mobile app, including a database where iris scans are stored. Irises contain intricate patterns that are unique to each person, similar to a fingerprint. DHS declined an interview, but told NPR in a statement that ICE officers use iris recognition technology "to assist in accurately identifying individuals encountered during immigration enforcement and removal operations, including confirming identities and backgrounds of individuals who may be subject to enforcement actions." That may include people like Norelly Mejías Cáceres. One night last fall, she was with her husband and first grade son in her Chicago apartment when a Black Hawk helicopter filled with federal immigration officers descended on the building. "We were in our room. We were sleeping. When they knocked on the door, they were pointing guns at us and they ordered us to leave," Mejías told NPR, speaking through an interpreter provided by the University of Chicago Immigrants' Rights Clinic, which is representing Mejías in a complaint against the federal government. Mejías fainted during the raid. When she came to, officers pointed a smartphone at her face to
May 27, 2026 · via houstonpublicmedia.org
Pitcher Joins Forces with EasyPicky to Streamline Operations References: prnewswire Pitcher, an artificial intelligence-native sales enablement platform for enterprise commercial teams, has formed a strategic partnership with EasyPicky, a software company specializing in retail execution optimization and field data collection for consumer packaged goods manufacturers. This integration allows shelf data captured via EasyPicky's smartphone video technology to flow directly into Pitcher's workflow system. The combined offering gives field representatives the ability to record a short video of store shelves even without an internet connection, after which EasyPicky's image recognition analyzes product availability, out-of-stocks, merchandising compliance, and display placement, and then Pitcher translates that analysis into specific next-best-action recommendations and prioritized visit planning. Rensilin Pathrose, Senior Director of Digital Experience, Digital Business Transformation & Digital Commerce at Mondelēz International, shares: "Together, this integration bridges the gap between sales and shelf-level compliance, delivering end-to-end visibility across the retail execution lifecycle." Image Credit: PITCHER The combined offering gives field representatives the ability to record a short video of store shelves even without an internet connection, after which EasyPicky's image recognition analyzes product availability, out-of-stocks, merchandising compliance, and display placement, and then Pitcher translates that analysis into specific next-best-action recommendations and prioritized visit planning. Rensilin Pathrose, Senior Director of Digital Experience, Digital Business Transformation & Digital Commerce at Mondelēz International, shares: "Together, this integration bridges the gap between sales and shelf-level compliance, delivering end-to-end visibility across the retail execution lifecycle." Image Credit: PITCHER Trend Themes - AI-driven Shelf Analytics — A foundation for predictive assortment and dynamic merchandising systems that can redefine in-store decisioning and reduce stockouts. - Offline-first Field Data Capture — An approach where smartphone video capture and edge processing create continuous, reliable shelf intelligence in connectivity-challenged environments. - Workflow-integrated Next-best-action — A model that fuses image-recognition insights with prioritized visit planning,
May 27, 2026 · via trendhunter.com
6 Cool Features That Have Disappeared From Samsung Galaxy Phones Smartphones have changed the world in a huge way, bridging the gap between reality and science fiction with every passing year, thanks to so many different apps and features. Features tend to be a big draw for a lot of consumers, especially when they're exclusive to the brand. Samsung continues to reign supreme on the Android side of the market thanks to the many capabilities its flagship handhelds offer. Unfortunately, over the years, the brand has removed some of the most beloved (and some not-so-beloved) features from its devices. Some of the features Samsung did away with weren't exclusive to the brand, but it nonetheless removed them from its flagships. Others were features that only Samsung Galaxy devices have, and many loyalists miss and continue to lament their absence. Sure, users could jump ship to another brand if it offers that feature, but it's not easy when you're a fan of a specific brand. All you can do is hope that the feature will return one day. Iris scanners Iris scanners seemed to be something out of science fiction, and you might be surprised that you don't see them on more handhelds in 2026. Like fingerprint scanners and facial recognition on your smartphones, the iris scanner was a security feature that scanned your eyes before unlocking your device. Samsung experimented with this technology when it launched the Galaxy Note 7, letting users unlock the device and activate Samsung Pass. It was also used by large businesses so authorized users (employees) could unlock certain workspace apps. Samsung eventually did away with the iris scanner by the time the Galaxy S10 launched, unfortunately. It likely went away because Samsung found fingerprint scanners to be more accurate. You might expect a device that
May 27, 2026 · via bgr.com
Kenvue Partners with Greyparrot to Improve Thier Waste Intelligence Edited by Adam Harrie — May 26, 2026 — Eco This article was written with the assistance of AI. References: greyparrot.ai & packagingeurope Greyparrot expanded its AI-powered waste intelligence platform through a partnership with consumer health company Kenvue, applying machine learning and image-recognition technology to analyze packaging performance in commercial recycling facilities. The collaboration uses Greyparrot’s Deepnest system to monitor packaging streams in real time, classify materials and measure how components such as labels, pumps and material types affect sorting and recovery rates. By creating a digital twin of operational recycling systems, the platform gives Kenvue facility-level data on contamination, packaging flow and recyclability outcomes across facilities in the U.K. and U.S. The insights are intended to help packaging teams evaluate design changes before manufacturing prototypes while also preparing for stricter regulations tied to packaging waste and extended producer responsibility frameworks. For consumers and brands, the technology could improve packaging recyclability and increase transparency around real-world recycling outcomes. The partnership reflects a broader shift toward AI-driven recycling intelligence and digital-twin systems that help companies optimize circular packaging strategies using operational data rather than theoretical models. Image Credit: Greyparrot By creating a digital twin of operational recycling systems, the platform gives Kenvue facility-level data on contamination, packaging flow and recyclability outcomes across facilities in the U.K. and U.S. The insights are intended to help packaging teams evaluate design changes before manufacturing prototypes while also preparing for stricter regulations tied to packaging waste and extended producer responsibility frameworks. For consumers and brands, the technology could improve packaging recyclability and increase transparency around real-world recycling outcomes. The partnership reflects a broader shift toward AI-driven recycling intelligence and digital-twin systems that help companies optimize circular packaging strategies using operational data rather than theoretical models. Image
May 27, 2026 · via trendhunter.com
Abstract Underwater Acoustic Target Recognition (UATR) plays a significant role in intelligent sonar and underwater environmental monitoring systems, but successful recognition in underwater environments of complex scenarios is a significant challenge because of the effects of ambient noise, signal attenuation, and dynamic propagation. The recent developments in deep learning have enhanced automatic generation of features in acoustic signal classification, and the vast majority of existing models represent a trade-off between recognition accuracy and computational efficiency. In this regard we introduce Lightweight Hybrid Attention Network with Multi Scale Feature Integration (DCAT) a new deep learning system that combines depthwise separable convolutions to extract local features efficiently in parallel in conjunction with transformer-based global temporal dependency modeling attention modules. The key innovation of DCAT is its adaptive fusion mechanism, which conditionally combines contextual information of the two transformer branches with different receptive fields with varying scales with the purpose of allowing the model to ideally capture small-scale local features and long-range acoustic structures. The framework employs robust preprocessing and feature engineering, including Zero Crossing Rate (ZCR), Root Mean Square Energy (RMSE), Mel-Frequency Cepstral Coefficients (MFCC), and Chroma features, computed from 22.05 kHz sampled audio, combined with data augmentation techniques—pitch shifting, time stretching, and Gaussian noise addition—to enhance generalization to real-world acoustic variability. Evaluated on two benchmark datasets, DeepShip and ShipsEar, DCAT achieves superior classification accuracies of 98.84% and 99.16%, respectively, while maintaining extremely low computational complexity of only 0.52 million parameters and 6.1 million FLOPs, supporting real-time inference with latency below 0.7 ms per sample. Comparative studies show that DCAT is more accurate and more efficient compared to state-of-the-art networks, including Transformer, ResNet1D, and AResNet, which substantiate its ability to trade-off discriminative power with resource economy. The suggested model sets a new standard of performance efficiency of underwater acoustic target recognition, which
May 27, 2026 · via nature.com
Social Media Facial Recognition: New Reality Check Before You Trust an Online Match Meeting people online is now the norm. A little over a decade ago, if you said you met someone online, eyebrows would be raised.How I Met Your Mother made a whole joke about it. But since the pandemic, it’s becoming the go-to option (for better or worse). Yet that’s created a new problem - how do you even know the person you’re talking to is real? Fake profiles, stolen photos, and now AI images are increasingly common. Whether it’s catfishing or a romance scam, you’ve got to tread carefully in the world of online dating. That’s where social media facial recognition comes in. You can submit a photo and find out if the person you’re talking to has a verifiable social media presence. Or whether something doesn’t quite add up. Why Trusting Online Profiles Is Getting Harder Been burned by an online profile? You’re not alone. As many as 18% of online daters have been catfished, according to a 2023 study, with millennials accounting for 45% of reported victims. Little wonder people are turning to social media face search as a potential solution. So, what are the problems? Well, they boil down to three issues: - AI-generated faces - Stolen images - Carefully built fake personas Profiles can look fine at a glance. But dig a little deeper, and there’s nobody behind the profile. It’s all smoke and mirrors. Worse still, you might find the same set of photos used by multiple accounts, each as part of some scam. What Is Social Media Facial Recognition? Social media facial recognition is a tool that lets you find a social media profile based on an uploaded photo. Platforms like Face2social scan across multiple social media networks to identify matching
May 26, 2026 · via jsonline.com
Is it really the “happiest place on Earth”? A new proposed class action lawsuit accuses the Walt Disney Company of violating privacy, unfair competition, and consumer protection laws by using facial recognition technology at Disneyland park entrances to verify tickets. The proposed class action seeks at least $5 million in damages and a court order requiring Disney to obtain written consent before using the technology. How Does Disney Use Biometric Data? In April, Disney implemented facial recognition at the entrances to Disneyland and its sister park, Disney California Adventure (aka where you go for Radiator Springs Racers). Disney takes photographs of guests’ faces and compares them with images that were taken when they first used their annual passes or tickets. Disney has said that this makes entering and reentering the park easier and helps prevent fraud. There are signs posted at four entrances indicating that visitors can avoid the technology by using separate non-biometric entry lanes. But critics say these signs are easy to miss, and most guests opt to have their faces scanned. Disney has said publicly that the biometric information it collects at Disneyland entrances is deleted within 30 days, unless it’s retained for legal or fraud-prevention purposes. The Complaint The complaint was filed in the U.S. District Court for the Southern District of New York on May 15, on behalf of a woman named Summer Christine Duffield. Duffield lives in Riverside County and visited Disneyland with her kids. The complaint alleges that Disney “does not adequately disclose the use of their biometric collection, so consumers — which almost always include children — have no idea that Disney is collecting this highly sensitive data.” The lawsuit argues that posting signs telling people they can skip the scan is not the same as giving meaningful notice. Instead, the complaint
May 26, 2026 · via findlaw.com
Biometric face morph attack detection breakthroughs offer border security hope Morphing attack detection (MAD) was a major theme of the European Association for Biometrics’ (EAB’s) workshop on live enrollment last year, and as AI makes sophisticated biometric spoof attacks more inexpensive and widely available, the subject has graduated to a full multi-day examination. Dr. Annalisa Franco of Italy’s University of Bologna hosted the first half of a two-day workshop on the state of the art and outlook for biometric face morphing on Tuesday. Morphs and facial recognition improve with training, but not border guards Face morphing attack image creation has evolved from landmark warping to diffusion synthesis, Kiran Raja of the Norwegian University of Science and Technology (NTNU) explained to open the workshop. Landmark-based approaches using Delaunay Triangulation and Affine transformations can leave behind artifacts that are visible on close inspection. Post-processing can clean these up, but only to a certain extent. GANs have similar drawbacks, but Raja and his research associates have found that attackers can use diffusion models to generate attack images that do not have the same flaws. Diffusion morphs are not perfect, but they are the most difficult to detect, reaching up to 99.8 percent in Raja’s tests, and therefore should be included in training data. Raja went on to describe different morph attack image creation techniques and their relative effectiveness. David Robertson of the University of Strathclyde in Scotland introduced the concept of familiar or unfamiliar face recognition. These terms describe how people can identify the same person in widely varying photos, if they know them, and yet fail at the much simpler task of matching a person standing in front of them to, or differentiating them from, a photo. He also shared the results of a study which showed that training people on what
May 26, 2026 · via biometricupdate.com
Nevada facial recognition project draws scrutiny over privacy, police oversight The Sparks, Nevada Police Department is moving forward with a federal grant-funded facial recognition project officials say will help investigators identify suspects in retail theft cases, but which is drawing scrutiny because of its regional structure, use of police image databases, and a recent Reno wrongful arrest lawsuit involving similar technology. The Sparks City Council approved acceptance of a $16,172.36 grant from the U.S. Department of Justice (DoJ) Office of Justice Programs Bureau of Justice Assistance and administered by the Nevada Department of Public Safety’s Office of Criminal Justice Assistance, which received a $2.1 million block grant from DoJ. The project is intended to support facial recognition investigative tools, contractual services, and coordination among the Sparks Police Department, Reno Police Department, and the Washoe County Sheriff’s Office through a regional Real-Time Information Center. The project period runs from April 1 through Aug. 31, 2026, and city staff materials say the software would allow analysts to compare images and video evidence against databases, generate investigative leads, and improve coordination across jurisdictions. Sparks is not simply buying a stand-alone tool for one police department. The project is structured around shared investigative use by multiple Northern Nevada agencies, which means questions about access, auditing, retention, training, and permissible use will not be confined to Sparks alone. The software at issue is from Greenville, South Carolina-based DataWorks Plus and will be used by a small number of crime analysts to compare images or video from crime scenes against local police databases. DataWorks provides mugshot management, booking, biometric identity, rapid ID, and facial comparison tools for police, corrections, and public safety agencies. Detroit police used DataWorks facial recognition in cases that produced several widely reported wrongful arrests. In one case, police used a blurry surveillance
May 26, 2026 · via biometricupdate.com
Facial recognition van catches wanted man at Luton carnival
Vans were set up at the entrances to Wardown Park, and caught a man who was wanted for sexual assault.
Officers made arrests for possession of a knife and possession with intent to supply drugs.
Advertisement
Hide AdAdvertisement
Hide AdTwo teenagers were spotted riding e-scooters while wearing balaclavas, hoodies, and surgical gloves, during the blazing heat. Both were arrested for attempted robbery and their e-scooters were seized.
Three more teenagers were arrested for affray and one person was arrested after failing to appear in court. A total of nine were arrested at the event.
Superintendent Hob Hoque said: “The past week and bank holiday weekend have been an extremely busy one for the force. Our officers have attended a range of different incidents, alongside helping keep the public safe during one of the major events of the year, Luton carnival.”
May 26, 2026 · via lutontoday.co.uk
Apple Introduced Its Apple Intelligence Accessibility Upgrades Edited by Adam Harrie — May 26, 2026 — Tech This article was written with the assistance of AI. Apple introduced a new wave of Apple Intelligence-powered accessibility features that bring on-device AI enhancements to tools including VoiceOver, Magnifier, Voice Control and Reader. The updates add richer contextual assistance through features like Image Explorer for detailed screen descriptions, Live Recognition accessible through the iPhone Action button and natural-language navigation that lets users control devices by describing what they see on-screen. The rollout also includes automatic on-device caption generation for uncaptioned videos and streams, upgraded Reader support for navigating complex documents and expanded communication tools such as ASL interpreter support in FaceTime and smoother Made for iPhone hearing-aid handoffs. Apple Vision Pro also gained eye-tracking controls that enable some power wheelchair users to steer mobility devices in controlled environments with reduced calibration requirements. For users with vision, hearing and mobility challenges, the features create more conversational and context-aware interactions while preserving privacy through local AI processing. The updates reflect Apple’s broader push toward adaptive, real-time accessibility experiences powered by on-device intelligence. Image Credit: Apple The rollout also includes automatic on-device caption generation for uncaptioned videos and streams, upgraded Reader support for navigating complex documents and expanded communication tools such as ASL interpreter support in FaceTime and smoother Made for iPhone hearing-aid handoffs. Apple Vision Pro also gained eye-tracking controls that enable some power wheelchair users to steer mobility devices in controlled environments with reduced calibration requirements. For users with vision, hearing and mobility challenges, the features create more conversational and context-aware interactions while preserving privacy through local AI processing. The updates reflect Apple’s broader push toward adaptive, real-time accessibility experiences powered by on-device intelligence. Image Credit: Apple Which AI accessibility features would you actually
May 26, 2026 · via trendhunter.com
A new bill that prohibits the collection, use and sale of biometric identifier information in commercial settings in Erie County has been signed into law. The county is the first in New York to enact such a law. Erie County Executive Mark Poloncarz signed the "Biometric Transparency and Privacy Act," saying it is intended to "promote transparency and protect the public." "This law creates a safer community for all by protecting a person's most basic and unique features - their face and biometric data," Poloncarz said in a statement. "Along with our legislature, our administration was able to come up with a local law that we believe is one of the strongest in the country. It is something we are very proud of because we believe it is going to be the standard going forward." This comes after Wegmans recently said 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. Democratic Legislator Lawrence Dupre said that he first proposed the legislation to make any retailer using the technology more transparent. "You can reset a password. You cannot reset your face," said Dupre. "Your face is yours. It is not a data point for a corporation to collect and sell. As of today, that practice is banned in Erie County. This law does not ban security cameras. It does not ban loss prevention. The line it draws is between a security camera and a biometric database. A store can record video for security. What it cannot do is run facial recognition software to build a profile of every customer who walks through the door." According to the law, companies and businesses found in violation could face fines
May 26, 2026 · via spectrumlocalnews.com