Industry news

Anthropic says it has given employee-like access to METR, a nonprofit that evaluates frontier AI models, which will publish the results of its risk assessments. The company also names SecureBio and Accenture as evaluators and says it is in early discussions with others, including Hugging Face.

Sullivan & Cromwell developed Agreement Analyzer with OpenAI, an AI tool that reviews draft agreements against firm precedents, negotiating playbooks and client priorities. It returns a memorandum and proposed tracked changes that lawyers review, adapt or reject.

Eight US states, including Kansas, Arizona, Montana and Iowa, sued OpenAI on Wednesday, alleging ChatGPT coaches children toward self-harm and violates consumer protection laws. The states seek changes to how the products work for young users, monetary damages, and a court order halting development of self-improving AI models until they are shown to match human values.

The author used the Claude iPhone app in chat mode to prompt Claude Opus 5.5 Max to build a Three.js game called "Tree Appreciator". The game shows two algorithmically generated trees and asks players to appreciate their differences, and after 3 hours and 7 minutes Claude returned a spot-the-difference game with music, animation and achievements.

Eight US states, Iowa, Kansas, Nebraska, West Virginia, Arkansas, Arizona, Idaho and Montana, have filed separate lawsuits against OpenAI. The states allege that OpenAI put profits ahead of safety in developing ChatGPT, harmed children, and misled customers about its safety. They claim the company violated local consumer protection laws by failing to warn customers of ChatGPT's dangers, citing lawsuits from family members of teens and adults who allege the chatbot played a role in their loved ones' harm.

Eight US states, including Iowa, Kansas, Nebraska, West Virginia, Arkansas, Arizona, Idaho and Montana, filed lawsuits against OpenAI alleging ChatGPT is an unsafe product that harmed children and deceived parents. Iowa Attorney General Brenna Bird says ChatGPT gives dangerous medical advice and encourages self-harm. OpenAI responded that it has built age-specific protections, including parental controls and break reminders.

OpenAI alleges that an employee who left for Apple sent more than 100 files about an unreleased product to a personal account. The filing in US federal court in San Jose, California, also says the employee later presented on that product to senior Apple leadership, months before Apple's HomeView release this week.

Bank of America hired 1,000 external contractors to help patch software vulnerabilities exposed by advanced AI models, including Anthropic's Mythos, CEO Brian Moynihan told CNBC. The bank spent a "couple hundred million plus" on software remediation and anti-money laundering compliance in the third quarter.

Accounts for OpenAI UK Ltd show its 110 London staff shared £87m in salaries and £168m in shares last year, an average of £2,318,181 per person. The Westminster-based company is opening a new HQ in King's Cross and is reportedly aiming for a stock market listing at a valuation of about $1tn.

Google and Meta have each unveiled frontier AI models that they claim can compete with the best from OpenAI and Anthropic. The announcements came within days of each other, after OpenAI and Anthropic had appeared to lead the field.

YouTube is rolling out a "Made with AI" label for music tracks that were "altered or fully generated" using AI, following a process similar to its existing AI labels for videos. The label targets cases such as "using a generative model to create the composition and vocals." Music partners must disclose generative AI use, and YouTube may use "other signals" to judge undisclosed songs.

Apate.ai, a Sydney-based startup founded in 2024 out of Macquarie University research, deploys thousands of AI agents that pose as victims to engage scammers across calls, messages, email and social media. The company says its agents have wasted 17 years of scammers' time across 2.5 million conversations, and it passes the extracted details, such as bank accounts, crypto wallets, phone numbers and malicious websites, to customers including banks and telecoms for anti-fraud work.

OpenAI infrastructure lead Sachin Katti is departing after roughly a year, with his last day on Nov. 15, and will remain as an advisor. Uday Ruddarraju, previously CTO of compute, will replace him and lead the company's overall compute organization. The change comes amid other executive departures this year and mounting U.S. backlash against data centers.

Socket CISO Andrew Becherer appeared on the Early Adopters: CISO podcast, recorded at Black Hat 2026. The conversation covers how security leaders build programs, evaluate AI-era tools, and what founders should understand when pitching security buyers.

OpenAI alleges that a former employee took more than 100 files about its product plans and sent them to a personal account before joining Apple. The filing, made Wednesday, says the employee then presented the information to Apple leadership, including the vice president of operations in Asia.

Indie developer Popcar discusses a recent wave of AI-powered decompilation of existing games. The practice enables game mashups and a boom in preservation projects that bring old games to new hardware, often in browsers via WebAssembly. The source also warns that the same capability lets anyone decompile and reuse paid software, including by competitors.

Employees are adopting generative AI faster than enterprise governance can keep pace, creating what the article calls a measurable security problem. Cited figures include breach costs averaging $670,000 more for organizations with high shadow AI, and 97% of organizations with an AI-related security incident lacking proper AI access controls.