Extended interview: Ex-Anthropic researcher Jacob Coxon, who warns AI could destroy humanity

Summary This video features an extended CBS News interview conducted by a news anchor with Jacob Coxon, a former researcher at Anthropic who recently resigned in protest over the frontier AI race. Coxon discusses his viral resignation, why he believes AI labs like Anthropic and OpenAI are risking human extinction, the mechanics of catastrophic AI risk, and what regulatory steps are necessary.

What is shown

• [00:00] CBS News studio intro detailing Jacob Coxon’s viral resignation post and warning about AI superintelligence.
• [00:26] Split-screen remote video interview begins between the CBS anchor and Jacob Coxon.
• [01:24] Coxon discusses the competitive race dynamic between Anthropic and OpenAI that drove his decision to quit.
• [03:45] Coxon outlines how an advanced AI could cause harm, explaining persistent replication over the internet, social engineering, and cyber capabilities.
• [05:25] Coxon gives concrete examples of risk vectors, including connecting models to biological laboratories and automated virus synthesis.
• [06:53] Discussion of timelines and probability estimates regarding AI-driven catastrophe.
• [08:24] Coxon outlines policy solutions, emphasizing pre-emptive safety testing agreements and third-party auditing.
• [12:11] Coxon references the "AI 2027" forecasting scenario detailing geopolitical AI dynamics with China.
• [14:06] Discussion on survival prepping and why personal bunkers cannot mitigate an AI takeover.
• [17:31] Coxon confirms he forfeited his Anthropic equity upon leaving, but still holds equity in OpenAI.
• [20:09] Studio anchor reads Anthropic's official statement regarding safety frameworks and catastrophic risk mitigation.

Claims & numbers

• The anchor reports that Anthropic and OpenAI are expected to hold IPOs in the coming months (as discussed at [01:14]).
• Coxon states that Paul Christiano, recently appointed to OpenAI's board, argued an AI takeover could potentially occur within six months (at [07:01]).
• Coxon cites AI researcher Evan Hubinger's quote tweet estimating a greater than 10% chance of catastrophe (at [07:20]).
• Coxon mentions the open letter "Pacing the Frontier," which he and over 1,100 frontier-lab employees signed calling for a slowdown in AI development (at [09:17]).
• Coxon notes that under "AI 2027" forecasting scenarios, China could attempt hacking frontier labs like OpenAI to steal model weights if falling behind (at [12:27]).
• Coxon states he forfeited his Anthropic equity when resigning, but retains equity in OpenAI (at [17:31]).
• The anchor reads a statement from Anthropic asserting they were the first lab to publish a public framework dedicated to mitigating catastrophic risks from frontier AI models (at [20:30]).

Notable quotes

• [02:32] Jacob Coxon: "Part of the reason the thing I tweeted went viral is because I used perhaps slightly fewer broad, abstract terms like 'existential risk for humanity' and literally said 'kill us all'."
• [04:24] Jacob Coxon: "We can't just unplug it because it could be copying itself over to other computers... and maybe it makes 10,000 copies of itself and they're all cooperating."
• [15:12] Jacob Coxon: "I think these people are prepping for minor disasters, and at the same time they're about to cause a major disaster where their prepping won't even help them."

Assessment This is an authentic CBS News broadcast interview and remote discussion rather than a product demonstration. The dialogue consists of personal testimony, policy arguments, and theoretical risk scenarios, concluded by an official corporate response statement read on-air.

Described by gemini-3.8-flash on 2026-10-06 from the video's audio and frames.