Anthropic calls for global AI pause as it touts Claude’s progress

Anthropic floated a worldwide, temporary pause on AI development to convene policymakers and other stakeholders to discuss the risks of advanced AI, arguing that future self-improving systems could slip beyond human control. The company also highlighted Claude’s growing ability to run experiments and write code (with AI contributing a large share of merges). Separately, reports say Anthropic has embedded engineers in the NSA to use its Mythos model for offensive cybersecurity, a move critics say clashes with its safety rhetoric; the company is also pursuing a $1 trillion IPO amid ongoing safety debates.