OpenAI engineer arrested for bugging execs, Character.AI goes bankrupt
An OpenAI engineer's arrest, Character.AI's bankruptcy, Gemini 4 Argon reaching subscribers, X's Grok and Cursor bundle, Anthropic ditching Git, and the labs' grip on outside safety testers.
An OpenAI engineer was arrested for bugging an executive meeting room Prosecutors say Abdulrahman Alfozan plugged a hidden recording device into the back of a TV in a conference room at OpenAI’s San Francisco office around 28 September. OpenAI’s security team found it two days later, and he was arrested and fired on Friday before appearing in court on Tuesday. OpenAI says it has no sign he got at model weights, source code or customer data. Source
Character.AI filed for bankruptcy with Disney offering $20m for some assets It filed on Tuesday, with Disney as the opening bidder, offering $20m plus a licence to its technology. Character.AI keeps running during the sale, but chief executive Colin Adams says the platform will be wound down if no bid comes in for the rest. Source
Google’s Gemini 4 Argon is out for Pro and Ultra subscribers It went into the Gemini app on Tuesday, and Google says it answers more directly, with less hedging and repetition. The coding tools around it, including Antigravity and AI Studio, need the Ultra plan. In the API, the introductory price of $2 per million tokens in and $10 out runs until 30 November, then doubles. Source
X bundles Grok, Cursor and Premium in one plan from $8 It’s called The X Subscription and runs from a free tier up to $200 a month, with the $30 tier the cheapest that gets the newest Grok models. It started rolling out on Tuesday, and anyone on an existing X plan can keep it. Source
Anthropic ditched Git after its code changes hit 30,000 a day That’s up from about 2,500 a day in May, and Git became the bottleneck: the history of its main code repository alone is nearly 100GB. Anthropic’s replacement, Scion, took five months to move to and needs about 200 times less computing power than its old Git setup. It wrote it up on Tuesday, but says Scion isn’t a Git replacement for anyone else and hasn’t released it. Source
Anthropic and OpenAI wanted a say in outside safety testers’ reports Reuters reported on Tuesday that both labs pushed for control over what independent testers could publish, citing trade secrets, according to two people familiar with the talks. Last month Dario Amodei said testers should publish their main findings without the company editing them. Anthropic now says it signs off only on reports from testers new to its models, and OpenAI says it just cuts secret details and lets testers say when large cuts were made. Source