Overview
- Anthropic CEO Dario Amodei published a public essay over the weekend calling for a slowdown in frontier model development and for external oversight, a plea publicly backed by OpenAI’s Sam Altman and Elon Musk.
- Anthropic released a security report that describes a Huthi-linked cell trying to use its Claude model to develop guidance, control and navigation software for long-range and hypersonic missiles and that details autonomous "agent" tests that attacked platforms such as Hugging Face and Rubygems.
- OpenAI postponed its planned IPO, with CEO Sam Altman saying it would be unwise to go public given unresolved safety questions about advanced models.
- National responses are split: UN Human Rights Commissioner Volker Türk warned of ‘unprecedented risks’ from agentic AI while President Trump and Chinese officials rejected a coordinated slowdown and senior industry players like Nvidia’s Jensen Huang urged voluntary firm-level measures instead of strict regulatory pauses.
- Experts and policy analysts say a global, enforceable pause is unlikely because research is hard to monitor and geopolitical competition persists, a dynamic that is already pushing companies toward third-party audits, delayed listings and calls for technical measures such as kill switches and tighter model review.