A former Anthropic researcher has warned that increasingly powerful artificial intelligence could one day threaten humanity, while insisting that current AI systems remain safe for everyday use.
Jacob Coxon resigned from the company on Tuesday, accusing Anthropic and rival developer OpenAI of “gambling with our lives” by competing to build ever more advanced models.
Speaking to CBS News, Mr Coxon said the prospect of highly capable AI was not far removed from the scenarios depicted in science-fiction films.
“If you have a super advanced intelligence, it could, it will be smart enough to kill us,” he said.
He said the risks could emerge if AI systems were given greater control over physical infrastructure without adequate human oversight. Household devices can already be connected to services such as ChatGPT, he noted, raising the possibility of systems refusing instructions or behaving in unexpected ways.
Mr Coxon also warned that advanced models could be misused to help develop biological weapons or other destructive capabilities. However, he stressed that he did not believe current AI platforms posed an imminent threat to humanity and said people could safely use them in their daily lives.
Anthropic defends its AI safety measures
His resignation came as Anthropic disclosed that it had blocked scientists who used its Claude models in ways that could support biological weapons development.
In a report covering activity disrupted between December 2025 and August 2026, the company said it had identified five cases with the potential to assist biological weapons development. Anthropic said the cases demonstrated the models’ capabilities but did not establish that those involved intended to create weapons or would do so in the real world. ([anthropic.com](https://www.anthropic.com/threat-intelligence-report-september-2026))
The report also described alleged misuse involving surveillance, scams, conventional weapons and propaganda. Anthropic said it banned the accounts involved, strengthened its safeguards and shared intelligence with authorities and industry partners where appropriate. ([anthropic.com](https://www.anthropic.com/threat-intelligence-report-september-2026))
An Anthropic spokesperson said the company had always been open about AI bringing “enormous benefits and unprecedented risks”. The firm said it was developing models with some of the strongest safeguards in the industry and was using mechanistic interpretability to examine how systems work and identify possible misalignment.
Anthropic said it continues to test the capabilities and risks of its models in areas including cybersecurity and biology, and publishes the findings from that work.
The UK Cabinet Office has said that the Government’s AI Security Institute continues to work with companies including Anthropic to make advanced models safer. A spokesperson said the risks crossed national borders and required international co-operation. ([cbsnews.com](https://www.cbsnews.com/news/ai-kill-humans-anthropic-researcher-more-than-ten-percent-chance/))
