AI Chiefs Call for a Slowdown—Critics Say They Can Hit the Brakes Themselves
AI Chiefs Call for a Slowdown—Critics Say They Can Hit the Brakes Themselves
Days before Dario Amodei’s intervention, former Anthropic researcher Jacob Coxon resigned, accusing leading labs of “gambling with our lives” in the pursuit of self-improving AI. The departure sharpened a safety debate that had largely played out inside the industry.
On Saturday, Anthropic’s chief executive made the case public. Amodei warned that increasingly capable, autonomous agent swarms could become a grave cyber threat within six to 12 months. His prescription was not a halt, but a deliberate deceleration: “We must slow the pace at which we improve the capabilities of AI models.”
His three-part plan begins with a unilateral Anthropic commitment to give outside evaluators permanent, employee-level access to its systems. It then calls for frontier labs in democratic countries to coordinate safety standards—with government help around antitrust barriers—before attempting narrower global cooperation, including with China.
The warning drew unusually fast backing from competitors and prominent industry figures. OpenAI’s Sam Altman said losing control of the future to AI was “unacceptable” and insisted that AI “must always serve people.” Elon Musk was blunter: “Dario is right.”
But the endorsement did not dissolve the central contradiction. David Sacks noted that Anthropic and OpenAI are, by market share, revenue growth and model capability, effectively a frontier-AI duopoly—and told them: “go ahead. You guys are the frontier.” A similar critique, amplified by Yann LeCun, argued OpenAI had “all the control in the world” and could simply switch off its machines if it chose.
Other critics objected to treating all AI work as equally dangerous. One widely shared response urged a change in priorities: less “GenAI autonomous agents without oversight,” more cancer-screening technology. The fight, then, is not only over whether to slow down, but who must move first—and what kind of AI deserves to keep racing.
Write a comment