Overview
- Anthropic publicly urged a coordinated slowdown on Thursday, saying that only a simultaneous, verifiable pause by leading developers would reduce the risk that AI outpaces safety work.
- The company says rapid progress risks ‘recursive self‑improvement,’ meaning systems could design more powerful successors and erode human control over their behaviour.
- Anthropic filed for an IPO this week while it fights a legal dispute with the U.S. Department of Defense after refusing unrestricted military use of its technology.
- Reporting based on anonymous sources says some Anthropic staff have worked on‑site with the NSA to deploy Mythos Preview for cyber tasks, a claim the company has not fully detailed and that complicates its public safety pitch.
- Experts and rivals are skeptical a true global moratorium is possible given heavy private investment, U.S.–China strategic rivalry, and recent moves by the U.S. government to review powerful AI models before release.