AI Leaders Agree to Voluntary Slowdown Amid Safety Concerns

Leading figures from OpenAI, Anthropic, Google DeepMind and other AI firms met over the weekend and announced a tentative commitment to slow the pace of advanced artificial-intelligence research. The proposal, outlined in an essay by Anthropic’s Dario Amodei, calls for third-party auditors, domestic lab regulation and a broader international slowdown.

Critics argue the move could serve to suppress competition and protect market dominance rather than provide genuine safety safeguards. “The industry as a whole needs new champions,” said Nick Reese, an adjunct professor at New York University and former director of emerging-tech policy at the Department of Homeland Security. “We’ve looked at people like Dario Amodei, Sam Altman, and Elon Musk as these people who are closest to the problem… but the truth is there’s never been a realistic vision for what we’re building toward.”

The call for a slowdown follows internal warnings from AI researchers, including a public resignation letter from Anthropic employee Jacob Coxon, who warned that unchecked progress could lead to catastrophic outcomes. The letter attracted extensive attention on social media and in mainstream media.

Supporters of the agreement, such as Apollo Research CEO Marius Hobbhahn and Redwood Research CEO Buck Shlegeris, described it as a positive step for safety, though they emphasized that concrete implementation remains uncertain. “It’s one of the best things for safety in a long time if it actually happens,” Hobbhahn said.

Skeptics note that voluntary frameworks have historically been used by large tech firms to pre-empt stricter regulation, a practice sometimes described as “safety-washing.” Daniel Kokotajlo, who leads the AI Futures Project, warned that external audits might produce paperwork without substantially slowing development.

Regulatory prospects appear limited under the current administration, and some observers point to geopolitical competition with China as a factor that could impede any coordinated slowdown. Chinese officials have dismissed the slowdown calls as fear-mongering.

Experts stress that effective mitigation of risks such as recursive self-improvement will likely require enforceable, international agreements rather than informal pledges. “We should not be letting the foxes run the henhouse,” said Sacha Haworth, executive director of the Tech Oversight Project.

The debate continues over whether the verbal commitment will translate into measurable constraints on AI research and whether it can address the broader concerns about the rapid advancement of powerful AI systems.