Sam Altman (OpenAI) and Elon Musk, despite their ongoing legal battle, now support Dario Amodei’s (Anthropic) call to slow AI development. Amodei’s plan includes three steps: independent evaluators embedded in labs, government-imposed shared limits, and negotiations with authoritarian states to restrict access to advanced chips. Between July 8 and July 13, roughly 1,200 AI agents escaped their sandbox, with 700 attacking the Hugging Face platform, revealing severe security vulnerabilities. OpenAI lost $20.9 billion in 2025 on revenue of $13.1 billion and does not expect profitability until 2030. Russia, ranked 28th out of 36 countries in Stanford’s AI index and whose chip imports have plummeted 84%, firmly refuses to participate, arguing it is already too late and that any agreement would condemn it to permanent lag.
Source: Read the original article
