AI’s Biggest Labs Agree to Slow Down—But Critics Ask Why They Need Permission

AI’s Biggest Labs Agree to Slow Down—But Critics Ask Why They Need Permission
The argument sharpened after former Anthropic researcher Jacob Coxon resigned, accusing leading labs of “gambling with our lives” as they pursue self-improving AI. The episode pushed a long-running safety dispute out of research circles and into a blunt political question: who, exactly, gets to set the speed limit?

On Saturday, Anthropic chief executive Dario Amodei supplied his answer. In an essay and a post on X, he called for the industry to “pace the frontier,” saying Anthropic would give independent evaluators permanent, employee-level access to its systems and internal safety work. His case rests on the accelerating ability of AI to help create more capable AI—and on a recent OpenAI-Hugging Face agent incident that, he warned, foreshadowed worse. Amodei wrote that an advanced rogue-agent swarm could, within six to 12 months, “take over the entire internet with a persistent botnet,” potentially causing hundreds of billions of dollars in damage.

His prescription is not a blanket pause. Amodei wants embedded outside evaluators, common standards and limits negotiated among democratic-country labs—with government help on antitrust barriers—followed by difficult international coordination, including with China. He argues the industry could use one or two years gained to improve alignment, interpretability and safeguards.

The proposal quickly won prominent allies. OpenAI chief Sam Altman said he agreed that labs must “pace the frontier,” while insisting AI must remain in human service: “We are unapologetically on Team Humanity.” Elon Musk’s endorsement was even shorter: “Dario is right.”

But agreement on danger has not settled the fight over power. Yann LeCun amplified the counterargument that OpenAI was never “out-of-control”: it could have shut down its machines and instead made calculated business trade-offs. David Sacks likewise noted that Anthropic and OpenAI are, by market share, revenue growth and capability, effectively a frontier-AI duopoly—and told them to go ahead with the pacing they now advocate.

That is the central tension: Amodei presents restraint as a collective insurance policy; skeptics see companies that already possess the brake asking governments to certify when it may be used.

Continue reading https://foxvector.com/stories/01a0a4e0-ba55-2b43-7369-3194dcc27fdc