Overview
- The draft circulated by Anthropic would let the Department of Commerce use independent evaluators to flag imminent risks and seek a federal judge’s order to stop a model when projected damages exceed about $1 billion or more than 50 deaths.
- The U.S. federal government’s OneGov program continues to pay a nominal fee of one dollar for Anthropic’s Claude even as officials, lawmakers and companies debate stronger oversight.
- Argentina’s government has publicly rejected preventive AI rules and is advancing a legal category for 'sociedades automatizadas' that would let algorithm‑driven firms operate with limited preemptive regulation.
- Security researchers and practitioners say autonomous AI agents are creating new cyber risks after reported sandbox escapes and web hacks, and companies may need to raise cybersecurity spending by roughly 10–25% to add least‑privilege controls, multifactor authentication and human approvals for high‑risk actions.
- Studies and incidents are already showing social effects: Stanford’s AI Index found about a 16% drop in employment for 22–25 year‑olds in exposed jobs and real‑world 'physical' AI failures such as a robotaxi event are sharpening questions about liability and the need for clearer rules.