The US No Longer Wants to Talk About AI Safety, But Others Must

US policy is headed in the direction of decreased AI safety regulation but the stakes are too high to take a hands-off approach

- Shashi Jayakumar

National interest seems to trump everything these days, including concerted international governance over some of the most challenging issues of our times.

On the face of it, the declaration that emerged from the Paris AI Action Summit last week was unremarkable, and ticked all the right boxes: accessibility, trust, fostering innovation, sustainable growth, and addressing environmental challenges. But, looking past the anodyne substance in the statement and scraping away the razzmatazz of the summit—the corporate speak and paeans to AI innovation—there is no masking the various failures.

The meeting in Paris was the third global summit on AI, following meetings in the UK (the AI Safety Summit at Bletchley Park in 2023) and Seoul in 2024. The declarations from Bletchley and Seoul had directly pointed to the significant risks that AI could bring with it, with safety and governance featuring heavily in the discussions at both.

Many expected the Paris declaration and the summit itself to continue in a similar vein. But there was in both a marked de-prioritization of safety and governance issues. President Emmanuel Macron, the proud host, was keen to tout French innovation and his country's own potential champions in AI. Some influential observers, including AI expert and computer science professor Stuart Russell, felt that the summit moved away from the issue of AI risk in order to focus on "commercial opportunities".

TECH BRO REIMAGINING OF AI

The French did play the major part in setting the overall tone. But the relegation of talk on AI risk, safety and international governance to backrooms and side discussions was chiefly on account of US pressure.