AI news, 9 October: OpenAI’s revenue nearer $50bn than rumoured $70bn

OpenAI revenue run rate nearer $50bn. According to the Financial Times and confirmed by Reuters and Bloomberg, OpenAI has told investors its annualised revenue was approaching $50 billion at the end of September—about $20 billion below the $70 billion figure that circulated last month. The higher number apparently arose from investor attempts to compare OpenAI with Anthropic, which includes sales via cloud partners; OpenAI does not. The clarification weighed on AI-linked stocks, with Nvidia, Oracle and others falling as markets digested the more modest figure alongside still-strong growth claims.

Anthropic bans needless cruelty to Claude and launches cyber push. Anthropic updated its usage policy to prohibit “sustained and needless abusive or cruel behavior” toward its models, effective 12 November. The rule targets extreme, purposeless cases and does not cover ordinary frustration, dark fiction or testing; conversation-ending remains the main enforcement tool. The same update tightens rules on weapons software, surveillance and deceptive campaigns. Separately, Anthropic launched its Cyber Mission with 11 partners including CrowdStrike and Palo Alto Networks, plus free model-powered vulnerability scans for open-source projects that include exploit proofs and suggested fixes.

Fired OpenAI safety researchers publish open letter. Three safety researchers dismissed last week—Jasmine Wang, Tomek Korbak and Mikita Balesni—released a letter denying OpenAI’s claims of mishandling information or leaking. They warned the firings are creating a chilling effect that leaves colleagues afraid to raise concerns or collaborate with external safety groups, and urged the company to bring in promised outside auditors while keeping model reasoning monitorable. OpenAI maintains they were let go for a pattern of policy violations, not for raising safety issues.

Google unveils Gemini agent for enterprise work. At its Gemini at Work event, Google Cloud introduced a universal Gemini agent that takes high-level objectives rather than step-by-step instructions, plans tasks, uses tools and company data, and works across Gmail, Docs, Sheets, Calendar, Slack, Microsoft 365 and more. It can act as a “coworker” with its own email address. Currently in private preview for businesses, with broader availability expected soon; it can also run on Anthropic’s Claude models.

Manus parent raises over $500 million. Butterfly Effect, the company behind the AI agent startup Manus, closed a funding round of more than $500 million led by Boyu Capital and IDG Capital, with Tencent and others participating. The raise follows Chinese regulators blocking a reported $2 billion acquisition by Meta. A valuation around $4 billion has been mentioned but remains unconfirmed.