OpenAI Proposes 5% Government Stake; Anthropic Models Freed

OpenAI has proposed giving the U.S. government a 1-5% equity stake to seed a 'Public Wealth Fund' and ease political pressure, a plan discussed with the Trump administration for over a year. Separately, the U.S. lifted export controls on Anthropic's Fable 5 and Mythos 5 models after three weeks of restrictions, with Anthropic agreeing to enhanced safety protocols. Anthropic also launched Claude Sonnet 5, a mid-tier agent-focused model, and Claude Science, an AI workbench for researchers. Google released Nano Banana 2 Lite and Gemini Omni Flash for faster, cheaper image and video generation.

Written by AI from the sources it cites; it can be wrong — follow any citation to confirm.

12 sources2 min read

OpenAI offers US government up to 5% stake to ease political pressure OpenAI CEO Sam Altman has proposed giving the U.S. government a 1-5% equity stake in the company, worth up to $42.6 billion, to seed a 'Public Wealth Fund' that would distribute AI gains to citizens. The proposal, reported by the Financial Times, aims to defuse mounting Washington pressure and has been discussed with the Trump administration for over a year, though no final terms have been agreed. 1 2

US lifts export controls on Anthropic's Fable 5 and Mythos 5 models The U.S. government lifted export curbs on Anthropic's Claude Fable 5 and Mythos 5 models after three weeks of restrictions, following Anthropic's agreement to enhanced safety testing, red-teaming, and 24/7 jailbreak monitoring. Fable 5 is now available globally, while Mythos 5 access remains limited to vetted U.S. organizations through Project Glasswing, with broader access under negotiation. 3 4 5

Anthropic launches Claude Sonnet 5 for cheaper AI agent tasks Anthropic released Claude Sonnet 5, a mid-tier model designed to run autonomous AI agents at performance levels previously requiring larger, more expensive models. It is now the default for Free and Pro plan users, signaling a shift from chat-focused AI to agentic workflows. 6 7

Anthropic unveils Claude Science AI workbench for researchers Anthropic launched Claude Science, an AI workbench for scientists that integrates over 60 scientific databases, a reviewer agent for error checking, and compute management. Available in beta to all paid Claude subscribers, it runs on Claude Opus 4.8 and offers compute grants of up to $30,000 for researchers. 8 9

Google launches Nano Banana 2 Lite and Gemini Omni Flash models Google released Nano Banana 2 Lite, its fastest and most cost-efficient AI image generation model, and Gemini Omni Flash for video generation and editing. The models are available through Google AI Studio, the Gemini API, and consumer products like AI Mode in Search, targeting faster and more affordable generative media. 10 11

OpenAI's LLM harness solves nine theoretical CS problems Researchers used a prover-verifier pipeline pairing GPT-5.5 Pro with Claude Opus 4.8 to solve nine open problems in theoretical computer science and mathematics, including four from the Conference on Learning Theory. The team plans to extend the method to all scientific fields. 12