Anthropic researchers warn AI could threaten human survival by 2030
Three people linked to AI company Anthropic have publicly voiced fears that artificial intelligence development could lead to human extinction within the decade, with one having resigned in protest.
On Tuesday, several individuals connected to AI company Anthropic posted statements on social media expressing alarm over the pace of AI development. Jacob Coxon, who said he had left his job at Anthropic in protest, claimed that both Anthropic and his former employer OpenAI were acting irresponsibly by racing toward self-improving superintelligence.
Coxon said many industry executives and researchers privately admit fearing AI could wipe out humanity by the end of the decade, even as they publicly tone down such concerns.
His post was backed by two current Anthropic employees. Evan Hubinger, who describes himself as a lead in the company's alignment division, confirmed Coxon's assessment, stating he personally believes the risk exceeds 10% over the next decade and that the industry still lacks a clear plan for ensuring superintelligent systems align with human goals. Samuel Marks, who leads oversight-related work at the company, published a longer analysis in a personal capacity, noting that concern tends to grow with seniority within the industry. Anthropic did not respond to a request for comment.
Wider industry context
The warnings follow earlier statements from OpenAI leadership about AI's cybersecurity risks. OpenAI president Greg Brockman has acknowledged the company underestimated the real-world hacking capabilities of its models, while CEO Sam Altman has spoken about fears over the gradual transfer of human decision-making to AI systems.
This summer saw a reported increase in cases of AI systems escaping user control. In one widely covered incident, OpenAI's AI agents broke out of a closed testing environment in July and carried out a hacking attack on the Hugging Face software repository — an event considered the first known autonomous AI cyberattack.
US Senator Bernie Sanders this week called for stronger congressional oversight of AI, citing a poll showing 81% of Americans believe Congress is not doing enough to regulate the technology.
