Agreement on AI risk still leaves a fight over who writes and enforces the rules

The Information reports that Anthropic, OpenAI and Google discussed an industry safety body. But the public alignment among executives has not produced a shared legislative plan.

Semafor reports that OpenAI favors bipartisan Senate Commerce negotiations that Anthropic, some Democrats and safety groups worry could leave too much self-policing. In the House, Mike Johnson emphasizes partnership with industry. Lori Trahan wants Congress to remain in session to work on AI, while Democratic leaders are still discussing their framework. The disagreement is about authority and enforcement, not simply whether risks exist.

There is also a new European response. Reuters reports that Germany calls halting AI development unviable, emphasizing digital sovereignty and international cooperation involving both the US and China. Spain favors an international agreement; the UK says regulation should be evidence-driven.

Endorsements of “safety” tell us little until they specify who can inspect systems, publish adverse findings and impose consequences.

Synthesis: the data-retention dispute in this briefing illustrates a related tension. Safety monitoring itself can require access that customers resist. Effective governance must define rights and responsibilities precisely; general assurances cannot resolve competing interests.