Clearview AI has developed an experimental AI tool that could extend a facial recognition search into broader automated research on a person’s online identity, background, and associates.
The prototype, called InquiryIQ, can conduct web and image searches, browse webpages, and run Clearview facial recognition on photographs it encounters.
Clearview has increasingly positioned facial recognition not merely to identify an unknown person, but as a starting point for gathering information and mapping possible relationships around that person.
However, it shows that using facial recognition to move from identification to personal information and relationships has long been part of Clearview’s technical conception.
The company says it documents each facial recognition search, including the probe image, the search’s purpose, and the user’s identity.
Sep 10, 2026 · via biometricupdate.com
iOS 27 lists three changes to VoiceOver, Apple’s built-in screen reader: richer descriptions of images, Live Recognition, and Action button integration.
Apple’s published record for those three items runs to about a dozen words in total.
Apple’s iOS 27 wording is that those descriptions become richer, which points to more detail than a short label.
Live Recognition is a name without a description Live Recognition appears in Apple’s VoiceOver entry for iOS 27 and nowhere else that Apple has published.
Action button integration, and which iPhones have the button The third item ties VoiceOver to the Action button, the customizable button Apple confirms on iPhone 18 Pro and iPhone 18 Pro Max in its tech specs.
Sep 10, 2026 · via macobserver.com
We had just grabbed a quick breakfast when Michael Friedman unexpectedly joined us.
Yes, that Michael Friedman.
The "AP, Michael Friedman, my friend's cooler" rap lyric guy.
The A-theory and B-theory of time were concepts coined in 1966 by a philosopher and professor, Richard M. Gale.
The Pattern Recognition B-Theory Series 1 costs CHF 150,000 on strap and CHF 120,000 on bracelet.
Sep 10, 2026 · via hodinkee.com
Norway is considering tighter restrictions on smart glasses and other camera-equipped wearable technology as officials raise concerns about surveillance, facial recognition and personal privacy.
Karianne Tung, Norway's minister of digitalisation and public governance, said the government wanted stronger regulation as artificial intelligence is increasingly combined with cameras and microphones embedded in glasses, headphones and other everyday devices.
The proposals do not amount to an announced blanket ban on smart glasses.
One measure under consideration is prohibiting functions such as facial recognition when they are used to identify other people in public spaces.
Norway's government plans to establish an expert group to advise on regulation and privacy protections for wearable technology.
Sep 10, 2026 · via thedailystar.net
Starting in mid-2026, the Transportation Security Administration (TSA) expanded its biometric identity verification program, TSA PreCheck Touchless ID, to 65 domestic airports.
The 65-Airport Milestone for TSA PreCheck Touchless ID Expanding Biometric Infrastructure Nationwide The expansion of Touchless ID to 65 airports by mid-2026 scales biometric infrastructure across the U.S. national aviation network.
Many travelers routinely compare TSA PreCheck, Clear, and Touchless ID to evaluate expedited security options.
When entering the general PreCheck screening area, follow overhead directional signs or floor markings designated specifically for "Touchless ID" or "Biometric Verification."
Touchless ID builds upon this modernized digital identity framework, creating a seamless pathway through terminal screening.
Sep 10, 2026 · via thetravel.com
Overview YOLOv11n-face-detection is a lightweight face-detection model from AdamCodd, built on the YOLOv11 nano architecture and fine-tuned for 225 epochs on the WIDERFACE dataset. It detects faces in images through the Ultralytics YOLO framework and is distributed as a model.pt PyTorch checkpoint. The most important consideration is its target range: it performs well on the WIDERFACE evaluation categories for easy and medium examples, but performance drops on hard examples and the model card states that it works best with frontal or slightly angled faces larger than 20 pixels. The supplied information does not report parameter count, input resolution, training batch size, hardware, inference latency, or VRAM requirements, so those values should not be assumed. Best use cases Face detection in ordinary photographs. The model is suited to finding human faces in images with frontal or slightly angled views. Its WIDERFACE validation AP reaches 0.9420471677096086 on the Easy split and 0.9210357271019756 on the Medium split, which supports use in common portrait, group-photo, and camera-image detection workflows. Lightweight face-presence filtering. A small YOLOv11 nano-based detector can serve as a first-stage filter before a recognition, alignment, or image-processing pipeline. For example, an application can detect whether an image contains a face, crop detected regions, and pass those crops to a separate face-recognition or quality-assessment system. The model performs detection only; it does not identify people. Face cropping for downstream recognition. The detector can provide bounding boxes for a later recognition system. This is useful when the application needs to locate faces before embedding extraction or identity classification. The model card does not provide recognition capability, identity labels, landmarks, or embeddings, so those functions require another model or a fine-tuned downstream system. Moderate-scale crowd and surveillance-image detection. WIDERFACE includes difficult face-detection conditions, and the model reports a Hard validation AP of 0.8099848364072022. That result
Sep 10, 2026 · via hackernoon.com
Real-time facial recognition in public spaces is illegal in most places, including New York, where the movie is set, and even in private spaces LFR use is tightly regulated.
Particularly given the response of Spider-Man’s police contact to his facial recognition use: “I’m going to pretend I didn’t hear that.”
And the response of the law enforcement authorities using facial recognition on the Delhi protestors?
In the U.S., court records allege ICE officers used live facial recognition to identify people observing immigration enforcement interactions, and in at least one case revoke travel credentials.
Spider-Man, on the other hand, realizes that he has been tricked about who the “bad guy” is in the scenario in which he has illegally deployed live facial recognition in public.
Sep 9, 2026 · via biometricupdate.com
Concealing the use of unpopular spying tools isn’t anything particularly new for law enforcement—cops have been doing it for years—but it’s just as wrong now as it was 20 years ago.
This is especially troubling when many towns are signing contracts with Flock and other ALPR vendors with little to no public oversight.
This trend is certainly not lost on law enforcement agencies.
Hiding the fact that they’re using ALPRs from Flock and other vendors is one way of avoiding scrutiny and bad PR.
But law enforcement and their spy tech vendors keeping people in the dark about the surveillance technologies trained on them predates the Flock backlash by decades.
Sep 9, 2026 · via eff.org
About 8,200 Michigan residents received enhanced driver's licenses or state IDs bearing the same unfamiliar man's face and a generic birth date after a vendor printing error affected thousands of government-issued credentials.
The Michigan Department of State said the error involved a batch of enhanced licenses and IDs printed Aug. 26 by IDEMIA, the state's card vendor.
The image and birth date did not belong to another Michigan resident, according to the state.
Corrected cards have been mailed automatically, and affected residents do not need to request replacements.
Enhanced IDs and REAL IDs are not identical.
Sep 9, 2026 · via lawcommentary.com
In traditional farming, water use is typically high, and manual work is required on a regular basis, leading to inefficient use of resources and late crop disease detection.
In order to overcome these problems, this research suggests an IoT and artificial intelligence (AI) based smart hydroponics system for a sustainable agricultural system.
The application of AI image recognition for early disease detection and the real-time and historical plant data analysis for resource-management recommendations are suggested.
Hence, the study was proposed to develop the integrated smart hydroponics system which integrates AgriFlow, hydroponics, IoT, AI, remote access, and data driven optimization.
AgriFlow: An AI and IoT-Based Smart Hydroponic Farming System AgriFlow : An A IoT-Enabled Smart Hydroponic Farming System for Intelligent Monitoring, Automated Resource Management, and Plant Disease Detection AgriFlow: An AI and IoT-Based Smart Hydroponic Farming System 2.
Sep 9, 2026 · via verticalfarmdaily.com
Key takeaways from the Big Technology Podcast interview with Mistral AI CEO and co-founder Arthur Mensch (January 2026): Models are commoditising fast.
With models converging, Mistral's focus is on enterprise applications and orchestration — solving the friction enterprises actually run into rather than chasing a "god model."
Mensch dismisses AGI as "too simple a concept"; real problems require specialisation, not one system that solves everything.
Open-source models let organisations run on their own terms, customise with proprietary data, and ensure no foreign provider can shut them off.
Mistral was first to release open-source models and sees itself as "the western open-source provider," filling a gap left by US companies that don't invest in open source.
Sep 9, 2026 · via app.dealroom.co
However, estimating the carbohydrate content of meals remains difficult, particularly for locally consumed foods.
Carbo AI is an artificial intelligence-powered mobile application that uses photographic image recognition, computer vision, and machine learning to identify foods, estimate portion sizes, calculate carbohydrate content, and provide instant dietary guidance.
The project was led by the Human Nutrition and Dietetics Team in the Department of Health and Biomedical Sciences, Technical University of Kenya, in collaboration with experts in computer science, medical physics, and software engineering.
Carbo AI aims to improve carbohydrate literacy, strengthen diabetes self-management, enhance nutrition counselling, and demonstrate the potential of interdisciplinary AI-driven innovations to improve health outcomes.
For those joining in-person, lunch will be served after the seminar from 1:00pm.
Sep 9, 2026 · via lboro.ac.uk
The measure could pave the way for broader use of technologies like the one already tested at the Olympic Stadium.
Behind him is a photo of the stands at the Olimpico, showing the faces of Roma fans recognizable by their jerseys.
What We (Don’t) Know About the Software Installed at the Olimpico Seven years after the Sind test, in 2021, the facial recognition system at the center of this investigation was installed at the Olimpico.
He is also concerned that there is no way to verify what happened to the data that were collected.
The legal basis that the Olimpico facial recognition system previously lacked could now come from the legislative decree given final approval by the government on August 4, 2026.
Sep 9, 2026 · via irpimedia.irpi.eu
Rochester, N.Y. — The Monroe County Legislature unanimously passed a measure Tuesday aimed at protecting consumers' biometric information and requiring businesses to disclose when they use the technology.
"They want to know before they enter a store if that store is collecting their biometric data."
BACKGROUND: Monroe County weighs bill requiring biometric ID disclosures by retailers The legislation does not prohibit businesses from using biometric technology.
Stored biometric information will also have to be anonymized so it cannot reasonably identify a specific person on its own.
"This local law will protect consumers' biometric data and give people the power to choose not to shop at stores that collect it."
Sep 9, 2026 · via 13wham.com
Stores in Monroe County will have to disclose if they use facial recognition technology, according to a new bill unanimously approved Tuesday by the county Legislature.
The bill, introduced by County Legislators Rachel Barnhart and David Long, would require businesses that collect biometric data to post "conspicuous" signage notifying customers.
The data collection includes facial recognition, fingerprints, retina or iris scanning, or any other identifying characteristics.
The legislation stems from an incident in January, when a Wegmans store in New York City publicly posted signage stating that it used facial recognition software in its stores.
In May, Erie County became the first county in New York to place an outright ban on biometric data collection in stores.
Sep 9, 2026 · via wxxinews.org
The project builds on Weckert’s broader work using technology to expose how digital systems govern physical navigation and perception.
How Digital Camouflage confuses AI detection The shirt does not make its wearer invisible to a camera.
The Digital Camouflage design specifically targets key visual markers that human-detection models rely on to identify people.
His projects include Nonexistent, which uses an AI algorithm to generate synthetic portraits, and Machine Unlearning, a workshop exploring ways artists can manipulate and challenge AI systems.
It gives people a way to see an AI system fail at something that appears obvious to a human observer.
Sep 9, 2026 · via colombiaone.com
AI - On-PREM AMD's Threadripper Halo is a local-AI workstation for researchers with deep pocketsAI workstation promises to put up to 576 GB of HBM3e and 16 TB/s of memory bandwidth on your desk - ai and ml Hugging Face is too important to fall into Nvidia's hands$12.9 billion deal will inevitably cement Nvidia's market dominance and harm competition in the process. Regulators should take note - ai and ml Zuck's Muse to Spark joy with open weights release 'soon'While you wait, Meta says it’s taught the model to stop wasting tokens and ask for help a bit more often - ai and ml With Gemini 3.8 Flash, Google reminds everyone it's still in the raceAI model scores well, runs fast, and doesn't cost too much (yet) - security AI agents carried out every step of this ransomware attack – then left the victim an 80-page security auditAdding insult to injury Infosec - Security Russians are posing as Signal support to launch phishing attacksPLUS: US takes down Iranian propaganda sites; Marketing company asks 'Why Do We Have Your Information?' And more! - Security Microsoft patches failed to fix on-prem SharePoint, which is now under zero-day attackPLUS: China upgrades smartphone surveillance tools; Ring eases anti-snooping stance; and more - Black Hat and DEF CON DEF CON Franklin project enlists hackers to harden critical infrastructureVoting village reports have been so successful, says Jeff Moss, that the whole of DEF CON will now be included - Security EQT buys majority share in Swiss cybersecurity biz AcronisWent at equivalent of $3.5B+ valuation for entire firm, though portion sold not specified - Malware Month Ten years since the first corp ransomware, Mikko Hyppönen sees no end in sightOn the plus side, infosec's a good bet for a long, stable career FOSS - Feel peak Windows
Sep 8, 2026 · via theregister.com
Abstract Deep convolutional neural networks have become widely adopted for food image classification, but their deployment on resource constrained devices remains challenging due to computational overhead, model redundancy, and increased architectural complexity caused by non contributing feature channels. Existing pruning approaches mainly emphasize parameter reduction, with limited focus on identifying and eliminating layer wise redundant feature responses while preserving semantic interpretability. In addition, Explainable Artificial Intelligence (XAI) techniques are largely confined to post hoc visualization and are rarely utilized for model optimization and structured compression. To address these limitations, this work proposes a Grad-CAM (Gradient-weighted Class Activation Mapping) inspired channel importance criterion for structured pruning of our SE-DenseNet121 model for food classification. The proposed framework identifies redundant feature responses, prunes non performing channels, reduces model cost and complexity, and retains only the most informative representations while preserving Dense Block connectivity. Using this importance score, 160 of 896 transition layer channels (17.9%) were removed. The pruned model was further optimized using Knowledge Distillation and cross entropy fine tuning. Experimental results show that the proposed framework reduces the parameter count by 8.39% and MACs by 5.15%, while achieving an accuracy of 98.13% (98.61% with test-time augmentation). Across five independent runs, the model attains a mean accuracy of 98.56% ± 0.07%, which is adopted as the primary reported metric. McNemar’s test shows a statistically significant but modest divergence between the pruned and teacher models, confined to 1.9% of the test images. Applied zero-shot to 3,045 images from three independent public datasets, the pruned model reaches 86.21% accuracy against the teacher’s 87.78%. These results demonstrate that explainability guided pruning improves computational efficiency while maintaining comparable recognition accuracy and interpretability for food image recognition. Funding Open access funding provided by SRM University-AP. No external funding was received for this research. The authors acknowledge institutional support
Sep 8, 2026 · via nature.com
A few years ago, artificial intelligence began making its way into manufacturing through image-based data comparison. Not exclusively, but especially for quality control applications, neural networks have since been trained to detect deviations from target specifications by comparing a component's actual state with reference images. Today, generative AI based on large language models helps users solve typically well-defined problems step by step. In an iterative process, humans provide prompts and instructions until a satisfactory solution is reached. Agentic AI operates very differently. For complex problems, it autonomously selects the AI tools (agents) required and develops solutions... ...with minimal human intervention. Everyday chatbots such as ChatGPT already illustrate the transition toward agentic AI. Users see how their request is broken down into multiple steps, such as searching, analyzing, and summarizing. However, they do not see the logic by which the system selects and sequences the appropriate tools. Agentic AI is already demonstrating its strengths in software development, customer service, IT operations, knowledge management, and administrative business processes. It excels wherever complex tasks consist of multiple steps and require coordination across various systems. Customer Service: In customer service, agentic AI can autonomously process a large portion of customer requests. It can do more than answer questions; it can execute complete service workflows and make independent decisions within predefined rules. Software Development: Modern coding agents analyze requirements, write code, run tests, and iteratively identify and fix defects. Retail: In retail environments, well-orchestrated AI agents can support inventory management, demand forecasting, logistics, and product administration. These agents independently combine information from different systems, prepare operational decisions, and in some cases execute them. Orcha-3: Bringing Agentic AI into Industrial Manufacturing The Orcha-3 platform, developed at Fraunhofer IWU, transfers this approach to manufacturing engineering. Orcha-3 is designed to address questions that cannot easily be solved using
Sep 8, 2026 · via idw-online.de
Police expand use of facial recognition technology - Published Police have used live facial recognition (LFR) for the first time in Devon. The technology, which scans faces in crowds from a live camera feed and checks them against a watchlist of wanted people, was used in Exmouth town centre at the end of August, officers said. Sgt Angela Galasso, from Devon and Cornwall Police, said the technology was probably "scanning far more faces than a police officer ever could", adding: "It's a tool. One of the many in our armour which help us to do our job." Any people identified by LFR still needed to be verified by humans, police said. Human rights group Liberty has called for legislation to clarify how police forces can use such systems. Last month the police used the same technology near the Boardmasters festival site in Cornwall. Galasso said a camera van had been deployed in response to anti-social behaviour, violent incidents and shoplifting in the town. She said: "We've got technology on board which is running with a predetermined watchlist of people. There's about 8,000 on the list. "The alert comes up, you're scanning the picture held on our system and the live picture. You're looking at that similarity scoring. "If it falls within a certain point, we then ask our uniformed officers to engage with that person and a second verification happens." The BBC spoke to shoppers on The Strand in Exmouth, where the cameras were used. "If you've done nothing wrong, then you shouldn't be afraid of it," said Denise Newman, from Budleigh Salterton. Paul Miller, from Exmouth, said he was "in favour of facial recognition with strict safeguards". Police said images of people who were not of interest would be "automatically and permanently deleted within seconds" and the force would
Sep 8, 2026 · via bbc.co.uk