General & Chatbots · AI safety debate
He left the lab. Then sounded the alarm.
Jacob Coxon left Anthropic to share big warnings. Here are the facts, debates, and events.
Tools: Claude · ChatGPT
Coxon described a possible catastrophic trajectory—not a prediction that extinction is certain or a finding that present systems can cause it.
What happened—and what did not.
Jacob Coxon left Anthropic on September 8. Before this, he worked at OpenAI and Anthropic. He posted online about fast races between labs. He said labs build smart tools too fast. He claimed they rush ahead without enough safety. Coxon chose to quit his daily job. News reports do not say Anthropic fired him.
News stories call him a former worker today. Some writers call him a whistleblower on websites. That word is only a label from commentators. It is not an official legal title here. He spoke to the news about his choice.
AP, WIRED, CNN and Coxon’s own account consistently describe a voluntary departure.
The verifiable television interview used here aired on CNN’s Anderson Cooper 360. CNBC separately covered the story.
Coxon told WIRED that he had worked on pretraining research and described internal conversations in which colleagues used phrases such as “crunch time.” WIRED also reported his view that Anthropic was more safety-conscious than OpenAI, while he argued that competitive pressure could still force future trade-offs.
Watch the argument unfold.
CNN’s interview separates the near-term state of current models from Coxon’s concern about a future, recursively improving system.
External video · CNN on YouTube
“There is a very real possibility of catastrophic outcomes.”
Coxon’s claim is about a future capability threshold. In the CNN interview, he explicitly said current models do not pose an extinction risk and focused instead on the pace of improvement.
He pointed to advances in coding, mathematics and cyber operations as evidence that capability growth deserves close attention.
The core concern is whether increasingly autonomous systems can reliably remain under human direction.
Labs may want safeguards while fearing that a rival will advance faster, creating pressure against unilateral restraint.
Claims, corroboration and uncertainty.
Evan Hubinger studies safety at Anthropic now. CNN and WIRED shared his personal thoughts. He shares the broad worry about smart machines. He guessed a greater-than-10-percent chance of human death within a decade. That is a personal guess, not proven fact.
AP reported two Anthropic workers agreed with him. Many people in this field care about this. Still, it does not settle the scientific debate. Guesses about future tools rely on unknown things. They depend on future power and new defenses.
Coxon pointed to online computer attacks as warnings. Anthropic shows public web pages about safety tests. Those web pages show tests on real cyberattacks. They show that people test for dangers today. They do not prove huge harm comes now.
What would a guarded future require?
Anthropic shared statements with CNN and WIRED. The company noted big benefits and new risks. It tests for dangerous skills inside its models. It supports legal pacing deals between tech labs. These rules check when strong systems get released.
Coxon suggested a starting deal among top labs. They would pause self-rewriting tools right away. Next, nations around the world would join in. This idea brings up very hard questions. Leaders must learn how to track rules fairly. Useful safety work must still move forward.
Test dangerous capabilities before release, with independent scrutiny and published methods.
Limit permissions, isolate sensitive tools, preserve human approval and support rollback.
Report failures consistently enough for researchers and regulators to learn from them.
Turn voluntary promises into rules that can be observed, audited and enforced.
The warning became two stories.
A researcher gave up a prestigious role to argue that competitive incentives were outrunning safety planning.
A dramatic extinction claim became a political and cultural symbol, often stripped of its uncertainty and technical conditions.
Read the underlying record.
This article was independently assembled from a televised transcript, Coxon’s published interview, wire reporting, secondary coverage, and Anthropic’s primary research materials. Roseram did not independently validate Coxon’s private conversations or calculate an extinction probability.
PRACTICAL NEXT STEPS
