An AI researcher who spent three years at two of the world's most advanced AI labs has quit, warning that neither OpenAI nor Anthropic is moving responsibly enough as they race toward superintelligent systems. Jacob Coxon, 27, announced his resignation from Anthropic in a series of posts on social media, declaring that the competitive pressure between U.S. AI companies and Chinese rivals is forcing labs to prioritize speed over safety.

Why Are AI Researchers Leaving Over Safety Concerns?

Coxon's departure reflects a deepening tension within the AI industry: what happens when companies racing to build more autonomous and capable systems begin making safety compromises because none wants to be the one that slows down first? After working on pretraining research at both OpenAI and Anthropic, Coxon concluded that both organizations are "racing straight to self-improving superintelligence and gambling with our lives".

Coxon

His warning carries weight because it comes from inside the labs themselves. Coxon says that researchers and executives at these companies take the possibility of superintelligence far more seriously than the public realizes. "The people building AI earnestly believe that it could kill us all by the end of the decade," he wrote, adding that "this is not a marketing stunt".

His

In interviews with the Wall Street Journal, Coxon went further, warning that some of the most aggressive scenarios could become possible as early as the end of 2027. He described colleagues using terms such as "crunch time" and "endgame" as AI companies race toward systems capable of increasingly autonomous behavior.

What's the Difference Between OpenAI and Anthropic's Approach?

According to Coxon, the two labs differ in how they understand the risks, but both face the same competitive trap. At OpenAI, he said, many researchers have not fully absorbed what he sees as the technology's potential civilizational consequences. At Anthropic, the risks are better understood, but executives and researchers remain trapped in a competitive dynamic: if they slow down, another company or another country may get there first.

This dynamic mirrors a classic technological arms race, except the thing being developed may eventually be able to improve itself. The concern is not theoretical. Industry leaders themselves have increasingly acknowledged the risks. OpenAI CEO Sam Altman has warned that cybersecurity failures involving increasingly capable AI systems could become serious without urgent action. Anthropic CEO Dario Amodei has repeatedly called for stronger safeguards and oversight as AI models become more powerful.

What Are the Key Tensions Driving This Debate?

Coxon's resignation highlights several competing pressures within the AI industry:

• Capability vs. Safety: AI companies are advancing capabilities at a rapid pace while safety research struggles to keep pace with the speed of development.
• Competition vs. Coordination: International competition, particularly with Chinese AI labs, creates pressure to move faster even when researchers recognize potential risks.
• Public Benefit vs. Existential Risk: AI could accelerate medicine, scientific discovery, education, and productivity, but it could also amplify cyberattacks, misinformation, labor disruption, and autonomous decision-making on an unprecedented scale.
• Individual Responsibility vs. Systemic Pressure: Researchers may want to prioritize safety, but the competitive structure of the industry makes slowing down feel like losing.

Coxon is not calling for AI development to end permanently. He says he remains optimistic that companies and governments can coordinate around safety, but believes stronger intervention may be required, including potentially temporary limits on further capability development.

How Can the AI Industry Address These Safety Concerns?

While Coxon's specific predictions about superintelligence timelines remain contested, his underlying question is harder to dismiss: if the companies building the most powerful technology in history believe slowing down could cost them the race, who decides when moving faster becomes too dangerous?

The challenge facing the industry includes several practical steps:

• Stronger Oversight Mechanisms: Establishing independent review boards and safety checkpoints that can slow development when risks are identified, without putting individual companies at a competitive disadvantage.
• International Coordination: Creating agreements between U.S., Chinese, and other AI labs to establish shared safety standards, reducing the pressure to race ahead unilaterally.
• Transparency About Risks: AI labs need to be more forthright with researchers, regulators, and the public about what they actually believe the technology could do and when.
• Funding for Safety Research: Dedicating substantial resources to understanding and mitigating risks, not just advancing capabilities.

Coxon's departure is far from the first time an AI researcher has left a major lab over concerns about whether development is moving too quickly. But his specific warnings about timelines and the language colleagues are using internally suggest that the urgency inside these labs may be higher than many outsiders realize.

The uncomfortable reality is that Coxon may ultimately be wrong about how quickly superintelligence arrives, or wrong about whether machines ever become powerful enough to escape meaningful human control. But the question underneath his warning is harder to wave away: as AI capabilities advance, the industry faces a genuine coordination problem that individual companies may not be able to solve alone.