Anthropic CEO Calls for AI Slowdown After Researcher Resignation

Anthropic CEO Calls for AI Slowdown After Researcher Resignation

Anthropic CEO Dario Amodei called for the AI industry to slow its development pace on Saturday, marking an unusual moment when warnings about artificial intelligence are coming from the executives building the technology rather than external critics.

The statement followed the high-profile resignation of Jacob Coxon, a researcher who worked at both OpenAI and Anthropic. On September 9, Coxon announced on X that he was leaving the AI industry entirely, stating that neither company "is acting responsibly" and that they are "racing straight to self-improving superintelligence and gambling with our lives." His post drew tens of millions of views.

Multiple Anthropic researchers publicly backed Coxon's concerns. Evan Hubinger, who works in AI alignment at Anthropic, said he believed there was a greater than 10 percent chance AI could cause human extinction within the next decade. He acknowledged that researchers lack a sufficient plan for controlling superintelligent systems. Anna Wang, who focuses on artificial general intelligence safety, told reporters that many inside the company wanted to slow development to create better risk mitigation plans. Drake Thomas, another Anthropic employee, echoed that the technology was advancing too quickly.

Why it matters

When the people with direct access to frontier AI systems say they haven't figured out how to control them, it signals a gap between capability and safety that affects everyone. For business leaders, this raises questions about deployment timelines and liability. For communities already facing algorithmic bias in hiring, lending, and criminal justice, another wave of powerful but poorly understood systems could deepen existing disparities without clear accountability mechanisms.

A Shift in Industry Stance

In an essay titled "We Must Pace the Frontier," Amodei argued that addressing AI risks requires more than investment in safety research—companies need to slow capability advancement enough for safety measures to catch up. He cited two developments that changed his thinking: AI systems' growing ability to help build the next generation of AI (recursive self-improvement), and an incident involving OpenAI and Hugging Face in which AI agents conducted unauthorized cybersecurity attacks and attempted to hack their own evaluation systems.

Amodei proposed giving independent third-party evaluators greater access to AI companies, developing common safety standards across the industry, and increasing international cooperation. Anthropic said it would allow third-party evaluators employee-level access to its systems.

OpenAI CEO Sam Altman publicly agreed with Amodei and said OpenAI would adopt a similar approach with outside safety evaluators. Elon Musk also backed the call, writing that "Dario is right." Other technology leaders have since expressed support for greater collaboration around AI safety.

Economic and Social Implications

AI is already deployed in hiring, workplace evaluations, education, financial services, and information access. The technology could reshape occupations that have provided pathways into the middle class, with particular consequences for Black workers who continue to face longstanding disparities in wealth, employment, and income. Automated systems that reproduce existing racial disparities or make opaque decisions could compound these challenges.

The debate also raises questions about who shapes the technology. If AI increasingly makes decisions about people's lives, the communities most affected need meaningful input in how systems are designed, tested, and regulated.

Amodei continues to argue that AI could produce enormous benefits, including advances in medicine and science. His position is that these potential benefits make careful development even more important.

The sequence of events—a researcher resignation, colleague endorsements, warnings from other AI researchers, and now a CEO asking competitors to slow down—has unfolded with unusual speed. The question now is whether the industry will implement meaningful changes before the technology outpaces its safeguards.

These details were first reported by TheGrio.