Leading AI executives have agreed that development needs stronger safeguards, but turning that agreement into an AI slowdown remains an unresolved challenge.
Dario Amodei, Sam Altman, Elon Musk and Demis Hassabis have backed a more cautious approach. But questions remain over what companies would actually limit and who would enforce the rules.
The Guardian’s analysis questions whether the industry’s safety promises will translate into action.
What Have Companies Committed To?
Anthropic’s Amodei proposed independent evaluators inside AI labs, shared safety standards among democratic countries and wider international cooperation.
OpenAI’s Altman committed to giving outside evaluators employee-level access. Hassabis supported the direction while saying the details needed work, and Musk offered a brief endorsement without outlining implementation steps.
These responses establish support for closer scrutiny, but no common timetable or measurable development limits.
Related: Top AI Leaders Call for Slowing Down AI Development
Who Would Write the Rules?
Critics worry that the biggest AI companies could shape oversight in ways that protect their own position.
Cohere CEO Aidan Gomez questioned whether dominant firms should set standards for the entire industry. Microsoft CEO Satya Nadella supported independent audits but called for broader representation.
The practical questions are significant: who selects the evaluators, how independent they are, and what happens when they identify a serious risk.
Washington Adds Another Obstacle
President Donald Trump has rejected slowdown calls, arguing that restrictions could help China catch up.
House Speaker Mike Johnson also opposed a development moratorium, although he supported independent audits and transparency, Reuters reported. That leaves some room for safety checks, while a government-backed pause faces resistance.
Related: Trump Rejects AI Slowdown Calls, Citing Competition With China
The next test is whether companies accept enforceable limits, meaningful outside scrutiny and consequences for unsafe development. Until those details are settled, public support for slowing down offers little certainty about how the AI race will change.
