AI rivals found rare agreement on safety. Putting it into practice is harder

The top names in artificial intelligence recently found common ground over safety measures. This surprising alignment garnered significant attention. Dario Amodei of Anthropic, Sam Altman of OpenAI, Elon Musk of SpaceXAI, and other influential figures in the industry expressed concern that the rapid development of AI needs to be slowed down. Their consensus emerged following a warning from an Anthropic researcher about the potential dangers the technology poses to humanity.

However, numerous hurdles stand in the way of a coordinated effort to regulate AI's progression. These include fierce domestic and global competition, the drive for profit, and pushback from President Donald Trump. Despite these obstacles, this moment may compel the industry to reconsider what constitutes success in AI. Nick Reese, a former director of emerging technology policy, suggested comparing AI success to that of airlines, emphasizing consistent on-time performance and safety records.

The leading AI startups could, in theory, regulate their own model development. However, technological advancements have been driven, in part, by the competition to create increasingly powerful models. This competition is backed by a White House that aims to maintain the U.S.'s technological edge over China. Additionally, Anthropic and OpenAI are vying for potentially record-breaking public stock offerings on Wall Street, although OpenAI has postponed its IPO this year to focus on safety efforts.

Proposals put forth by industry leaders require mutual collaboration and, to some extent, government regulation. Amodei recently detailed a plan to slow advanced AI development, which includes committing to ongoing, independent evaluations by outside teams. Anthropic has voluntarily agreed to this step, while OpenAI has expressed support. Altman welcomed a federal framework setting safety standards for advanced AI but did not believe waiting for legislation or an antitrust exemption was necessary.

Amodei also called for government regulation and international coordination, including with authoritarian regimes like China, to establish common safety standards and limits on unchecked AI progress. He suggested several levels of agreement, with the most feasible being a ban on certain narrow and dangerous AI uses. Achieving a full pacing or even a pause in AI development would be the most challenging, Amodei argued, comparing it to Cold War-era treaties that limited missile proliferation while preserving each country's deterrent capability.