A set of parents and their children in Illinois and California filed a lawsuit last week in federal court in Chicago alleging that Meta illegally used their Facebook and Instagram photos to build NameTag, an unreleased face-recognition system for its smart glasses, and to train generative AI models including Emu and Muse Image.

The proposed class action alleges that Meta violated Illinois and California privacy laws by extracting biometric information from people’s photos without notice or consent.

WIRED reported in June that code for NameTag had been secretly embedded in the Meta glasses AI companion app, which had been downloaded more than 50 million times. While the feature had not been enabled for users of the app, the analysis found that the system was designed to turn faces captured by the glasses into biometric signatures and compare them with so-called faceprints stored in a database on the user’s phone. That database was configured to receive updates from Meta. At the time, WIRED could not determine where the underlying faceprint data came from.

The complaint alleges that those faceprints are possibly derived from Facebook and Instagram images, citing reporting that Meta employees claimed that NameTag could recognize people through their Meta connections or public Instagram accounts, along with a company patent describing face matching against profile photos and other images held by Meta.

Meta told WIRED in June that it was “not building a central face database,” but it would not answer questions about whether NameTag would be opt-in or how the system would retain faceprints. The complaint acknowledges that Meta has not disclosed which images, if any, were used for generating biometric data, saying that information remains solely in the company’s possession.

The suit also targets Meta’s image-generation systems. Meta has said it trained Emu on large quantities of Facebook and Instagram images and text, with chief product officer Chris Cox calling those platforms a “data advantage” for its AI systems. The complaint alleges that the training process illegally harvested biometric information about people who appeared in the images. Muse Image, released this summer, had earlier drawn criticism after allowing users to generate images based on other people’s public Instagram accounts, a feature the company removed within days after saying it had “missed the mark.”

“This lawsuit is without merit and misrepresents our work. We’ve been transparent about how we use people’s information to build and improve our AI products. As for NameTags, nothing has shipped to consumers and no final decision has been made on what to do here, if anything,” a Meta spokesperson says in a statement. “If we do decide to roll something out, we will take a thoughtful approach and do so with full transparency. One decision we can be clear about—we are not building a universal face database.”

“People shouldn’t have to worry if their biometric information will be misused simply because their photographs appear on a social media platform,” Justin Boley, a partner at Wexler Boley & Elgersma and an attorney for the plaintiffs, said in a statement.

The plaintiffs are Francisco Alvarez and his son, both Illinois residents, and Jeremy Wahl, a California resident, and his 10-year-old daughter. But the proposed class includes people in Illinois, California, and across the United States whose images were uploaded to Facebook or Instagram or were submitted to Meta’s generative AI systems through prompts, dating back to September 4, 2021. The complaint estimates that the national class could number in the millions.

Under Illinois’ Biometric Information Privacy Act, the plaintiffs are seeking $5,000 for each intentional or reckless violation, or actual damages if greater, and $1,000 for each negligent violation, or actual damages if greater, as well as injunctive relief. The California claims seek additional damages and other relief.

What we know so far: Anthropic, the San Francisco-based AI company led by Dario Amodei, claims to have identified possible attempts by suspected state-sponsored researchers to exploit its Claude AI model to develop biological weapons. The company added that the users were blocked after repeated attempts triggered Claude’s built-in safety protocols.

Anthropic recently detailed at least five cases where suspected rogue actors “circumvented controls” specifically designed to prevent users in certain parts of the world from accessing Claude’s extensive knowledge base regarding infectious diseases, such as bird flu, and lethal venoms and toxins, like botulinum toxin.

However, the total number of attempts is much larger, with the company claiming to have identified about 35 distinct efforts between November 2025 and September 2026 by possible state-funded bioweapons developers to “obfuscate the purpose of their research to evade our safeguards.”

One of the suspected cases involved gain-of-function research into the transmission and immune evasion of the chikungunya virus. Another account was flagged for researching how the bird flu virus can be adapted to cause severe illness in mammals. Researchers were also banned for queries about orthopoxviruses – a genus of DNA viruses that includes variola, which causes smallpox, and mpox.

Anthropic acknowledged that it is difficult to determine whether research on toxins, poisons, and related substances is intended to aid the manufacture of bioweapons or to develop antidotes, vaccines, and other medical or therapeutic solutions. However, what caused the system to flag the aforementioned attempts was the use of various anonymizing techniques to evade Anthropic’s regional blocking.

In the aftermath of the bioweapons scare, Anthropic says it has upgraded its safety protocols to prevent threat actors from exploiting its AI models for nefarious purposes. “Out of an abundance of caution, we have launched recent models (most notably Claude Fable 5) with stronger safeguards that restrict access to a wide range of dual-use biological research queries,” the company stated.

Anthropic did not identify the regions from which these queries are believed to have originated, but the company is widely known to block users in China, Russia, Iran, North Korea, and several other countries from accessing Claude. However, it says it blocked all suspect accounts by May 2026 for violating its Supported Regions Policy.

According to Anthropic, misusing AI to create bioweapons is one of the “most serious risks” facing AI models. The company added that it incorporated the guardrails precisely to avoid situations like these, in which rogue nation-states could exploit AI to prepare deadly bioweapons and use them to wage biological warfare against their own people or Western democracies.

Flock automated license plate-reading cameras are rapidly popping up all over the US. In some areas, they’re also going down pretty quick, too.

Widespread opposition to Flock surveillance cameras has highlighted their inaccuracies and potential for abuse by authorities. A movement to “deflock” cities includes everything from canceling municipal contracts to covering cameras with trash bags and various forms of physical vandalism.

Now, if you’re cruising through the fictional town of Los Santos (modeled after Los Angeles) in a new Grand Theft Auto 5 “mod,” you could be tracked by any of the 235 Flock cameras there. But you can also cash in $600 for each one you shoot down.

The mod (a custom addition to the game) is called Grand Theft Automated License Plate Reader. Morry Kolman, the mod’s creator, said the in-game Flock cameras are based on “custom-built” 3D models, which subject players to “seemingly inescapable surveillance.”

Almost like living amid Big Brother, these virtual cameras take pictures of you and track your movements. If you commit crimes in the game — like speeding or hitting pedestrians — the cameras will report you to the police. They’ll even occasionally file “false reports” on you, just like the real cameras sometimes do. Players can also save the images captured by the cameras for use in social media posts or as wallpapers. 

And, because it’s Grand Theft Auto, you’re also allowed to fight back against the surveillance state, with access to weapons (like rocket launchers) that help you destroy the cameras. Once you’ve carried out an effective deflocking, you can pick up the mutilated remains and grab an in-game reward of $600 (the estimated value of the camera’s internal components based on a teardown conducted by Eyes Off Cedar Rapids). There’s also an integrated timer to keep track of how fast you deflock, allowing you to beat your PB.

A virtual protest

GTA 5 is chock-full of social commentary, from the dangers of consumerism to subtle critiques of extreme gamer behavior in characters like Trevor and Jimmy. The new Grand Theft Automated License Plate Reader mod builds on the general hatred toward and distrust of the massive Flock surveillance network adorning US roadways.

For some, it’s probably a fun (and less punitive) way to tap into unfulfilled vigilantism. But, according to Kolman, it also makes the game more difficult. “Not only does police surveillance become nearly impossible to evade, but representing Flock’s own history of inaccurate reporting, every time a player passes through a camera’s field of view, the mod creates a 1 in 20 chance that the ALPR will trigger a police chase without any cause,” according to the mod’s download page.

In many ways, the new mod is a reality check for real-world privacy. If having 235 cameras follow you around in GTA feels like an oppressive nightmare in a claustrophobic game world, just remember that the real Los Angeles County has more than 10 times that amount.

Joe is a freelance journalist. It all started with a long-running affection for building his own PCs, which he did for the first time as a teenager. It evolved into a lifelong enjoyment of putting words on the internet about the subject. He’s written for CNET, PCMag, Mashable and SlashGear as a freelance writer, and worked as a Senior Editor at Android Authority for 10 years. When he’s not writing about tech and science, he’s learning the ins and outs of DIY home repair, gaming, playing his bass guitars and posting help on PC building and gaming subreddits. He is a staunch believer that orange juice should have pulp.

See full bio

Book your exhibit table at TechCrunch Disrupt 2026 by September 18 at 11:59 p.m. PT. Tables are limited and can sell out before the deadline.

If you’re coming to TechCrunch Disrupt 2026 to find customers, meet investors, build partnerships, or put your product in front of the people who can move your business forward, don’t leave your presence to chance.

The exhibitor program gives your team a dedicated 6′ × 30″ table in the Expo Hall for all three days — a place to demo what you’re building, answer questions in real time, and turn passing traffic into meaningful conversations.

You’ll also get:

• Lead-generation tools through the Disrupt app so you can capture prospects and follow up after the event.
• 10 team passes so you can staff your table while your team takes meetings and explores the rest of Disrupt.
• Event-wide branding across the Disrupt website, app, and signage.
• Access to the TechCrunch Disrupt press list and a company profile on the event site and app.

Come ready to demo your product, start conversations, identify potential customers and partners, meet investors, and turn event traffic into relationships you can continue building after Disrupt.

Time is running out to secure an exhibit table at Disrupt — and once the remaining tables are gone, so is the opportunity to get your startup on the Expo Hall floor.

The real cost of waiting isn’t just missing a table. It’s missing the conversations, leads, and opportunities that could come from having your product in the room.

The deadline is September 18 at 11:59 p.m. PT — tables can sell out before then.

One week. Limited space. Don’t wait for the opportunity to disappear. Secure your exhibit table now.

Twenty-five leading mathematicians signed an open letter arguing that AI labs are threatening their intellectual work as they seek to one-up each other with solutions to famous math problems. Each signatory has been awarded the Fields Medal, considered the most prestigious prize in mathematics.

This week, NYU professor Tristan Buckmaster accused OpenAI of pressuring him not to credit a collaborator who works for Anthropic for solving an important math problem, and wondered if the company had used their work with Codex to produce its own groundbreaking proof over a marathon weekend of inference.

On Thursday, OpenAI withdrew its sponsorship of a math event at CalTech after the company was criticized by researchers at the university.

While the ability of AI models to solve the world’s outstanding mathematical challenges could be a boon to humanity, the signatories of the new letter argue that will only be the case if those solutions can be understood and communicated by the math community and, ultimately, the rest of the world.

“Often these solutions are announced in a rush, leaving no time for a proper writeup, the isolation of new methods and ideas, and citing relevant previous work of others,” they wrote — and OpenAI’s proof remains unverified. “As in all creative professions, this raises severe attribution and plagiarism questions. Moreover, without the willing mathematicians who must take care of their development and integration into the mathematical canon, AI-conceived ideas would never become fully alive and the crucial human transmission chain between mathematicians would be lost.”

With other mathematicians growing paranoid and wondering if their Codex use was in turn fed into OpenAI’s new models, there is real fear that the culture of open research will be threatened. Today, if frontier labs see a useful path to a discovery, they can spend tens of millions of dollars using LLMs to beat the original researchers to a proof — a dynamic that will incentivize secrecy.

This letter follows the Leiden Declaration, released by a working group of mathematicians in June. That document also grapples with the ways that LLM proofs will change their work and offers a set of recommendations for mathematicians, institutions, and policymakers.

As with software engineering and other areas where AI tools are changing workflows, mathematicians find a justification in the work around the work: The value in math isn’t just the proofs and who gets credit, but the intellectual super-structure that nourishes students, finds new questions and ideas, and integrates them into broader human civilization.

And if you don’t particularly care about the cutthroat world of high-stakes mathematical proofs, don’t forget: Your field of interest is next.

“The issues the mathematical community faces now are similar to issues that other scientific and creative professions are facing, and indicate issues that all of humanity might face: how to make sure that, as AI changes the way work is done, we do not lose sight of what that work was meant to achieve in the first place,” they wrote.

GitLab urged users on Thursday to patch their servers immediately against a maximum-severity path traversal vulnerability tracked as CVE-2026-85706.

The security flaw, discovered by a security researcher using the ‘s3ntago‘ handle and reported via GitLab’s HackerOne bug bounty program, stems from improper path confinement and missing authentication enforcement in the repository commits API.

Unauthenticated attackers can exploit CVE-2026-85706 “under certain conditions” to read arbitrary data (e.g., credentials, secrets, and sensitive information) from vulnerable servers.

While GitLab has yet to flag this flaw as exploited in the wild, one day later, cybersecurity company watchTowr reported that attackers have already begun searching for Internet-exposed GitLab servers unpatched against CVE-2026-85706.

“watchTowr Intel is already observing in-the-wild probes for the latest critical GitLab Path Traversal vulnerability, CVE-2026-85706, which allows attackers to read arbitrary files in a single HTTP request,” it warned.

“Based on recent GitLab vulnerabilities, we know the time until indiscriminate exploitation is likely not far away. [..] Defenders should also hunt through log files for HTTP POST requests to ‘/api/v4/projects/{id}/repository/commits/’ URIs containing ‘file.path’ parameters to identify potential exploitation attempts.”

Yesterday, GitLab patched a second critical vulnerability tracked as CVE-2026-87719 that stems from an insecure deserialization weakness in the GraphQL subscription serializer.

CVE-2026-87719 affects GitLab EE and allows authenticated users with Duo Chat access to steal sensitive credentials and Advanced Search instance configurations.

Admins warned to patch as soon as possible

GitLab fixed the two security issues in GitLab Community Edition (CE) and Enterprise Edition (EE) versions 19.3.2, 19.2.6, and 19.1 on Thursday, and urged users to patch their systems immediately.

“These versions contain important bug and security fixes, and we strongly recommend that all self-managed GitLab installations be upgraded to one of these versions immediately,” the company warned on Thursday. “GitLab.com is already running the patched version. GitLab Dedicated customers do not need to take action.”

In May 2023, GitLab addressed another maximum severity path traversal flaw (CVE-2023-2825) that exposes sensitive data, including proprietary software code, user credentials, tokens, and files on unpatched servers.

One year later, CISA and the FBI urged software companies to weed out path traversal security vulnerabilities from their products before shipping, saying that such flaws “have been called ‘unforgivable’ since at least 2007.”

More recently, in January, GitLab also patched a high-severity two-factor authentication bypass affecting community and enterprise editions that enables attackers who know the target’s account ID to circumvent two-factor authentication.

Since November 2021, the U.S. Cybersecurity and Infrastructure Security Agency (CISA) has flagged four GitLab vulnerabilities as exploited in attacks, including two (CVE-2021-22175 and CVE-2021-39935) in February this year.

The GitLab DevSecOps platform has more than 30 million registered users and is used by over 50% of Fortune 100 companies, including Nvidia, Airbus, T-Mobile, Lockheed Martin, Goldman Sachs, and UBS.

Update September 11, 09:39 EDT: Added watchTowr’s report of CVE-2026-85706 probing.

Join Mikko Hyppönen and security leaders from the NFL, CHANEL, and Atlassian for a two-hour digital summit on what AI-speed attacks change, what defenders should stop doing, and how to validate, decide, fix, and re-validate at machine speed.

Save your seat

AOC’s 16T20, priced at $49.99 (was $70), is a 15.6-inch Full HD IPS panel built to ride along with a laptop instead of staying on a desk. Resolution is 1920 by 1080 at 60Hz, response time is listed at 4ms, and the viewing angles are 178 degrees in both directions, so documents and video still look even when you share the screen across a table.

Official brightness is a reasonable 250 nits, which still looks decent in full sunshine. Meanwhile, the contrast ratio is equally respectable among its specifications, ranging from 800:1 to 1200:1. The display itself can show 262,000 colors, which is sufficient for common tasks such as emailing, spreadsheet work, video conferencing, and streaming. Measurements have revealed that this is better suited to interior use or under dim cabin lighting than the full-on glare of a bright patio.

Sale

AOC 16″ Portable Monitor FHD 60Hz USB-C IPS Ultra Slim Smart Cover 16T20

• EXPAND YOUR WORKSPACE ANYWHERE – Add a flexible 15.6-inch portable second screen to your laptop for smoother multitasking, enhanced productivity, and…
• FULL HD 1080P CLARITY FOR WORK & ENTERTAINMENT – Enjoy sharp detail with a 1920×1080 FHD panel, ideal for spreadsheets, video calls, gaming, and…
• IPS WIDE-ANGLE VIEWING WITHOUT COLOR SHIFT – The 170°/170° IPS display delivers accurate colors from any angle. Share the screen, review content…

This display excels in terms of size and weight. The display measures 13.9 inches wide by 8.8 inches high and is only 10 millimeters thick. Plus, it weighs only approximately 1.5 pounds, and that’s without the smart cover, which adds some weight but keeps the overall weight under 2.4 pounds. The cover folds into a kickstand for landscape or portrait viewing, and it also protects the screen when the display is put into a bag. If you desire to mount it on a wall later, the VESA 75 by 75 holes on the rear make it simple to set up.

On the back, there are two USB-C connections, which are configured so that they can be used for both video output and power via DisplayPort Alt Mode. This means you can run it from a laptop with a single wire, which is really convenient. There is also a Mini HDMI port for connecting to devices such as the Nintendo Switch, Steam Deck docks, and older PCs, but keep in mind that the box does not include a wall adapter for devices other than laptops, smartphones, or tablets. Power usage is also reasonable, with testers reporting around 4 watts at maximum brightness, which should not be too stressful on your laptop battery even during a lengthy flight or when working in a coffee shop.

The only minor drawback is the very basic audio, but it does the job with its 1-watt twin speakers, which will cover you for a conference or a streaming concert, as well as a handy 3.5mm headphone socket. There is no built-in USB hub or touch layer, so this is really a simple small supplementary display – which is probably precisely what you need. If you find yourself working late at nite, the on-screen menu includes a flicker-free backlight and a low-blue-light preset.