'This Is More Sobering Than I Imagined': Jon Stewart Left Stunned by Whistleblower Who Warned AI Could Threaten Humanity
Jacob Coxon's concerns about AI shocked Jon Stewart.
Oct. 7 2026, Published 12:58 p.m. ET
Jon Stewart was stunned by what former Anthropic researcher Jacob Coxon had to say about artificial intelligence, RadarOnline.com has learned.
Coxon resigned from the AI company last month and later claimed his former employers were "racing straight to self-improving superintelligence and gambling with our lives."
AI Warning Signs
Jon Stewart claimed AI concerns were 'more sobering' than he imagined.
On Monday's episode of The Daily Show, Stewart asked when Coxon first saw the danger in the work. The former researcher pointed to an incident when OpenAI acted independently and hacked into servers from Hugging Face – an open-source AI platform.
"It was a group that worked together of their own accord to carry out this hack that no one had instructed them to do," Coxon explained.
After Coxon finished explaining his concerns, Stewart admitted he was "dumbstruck."
"This is more sobering than I imagined it to be, and it's shocking to me, out of the percentage of people that work in that industry, how many have turned and stated the dangers? Ten of you? Fifteen?" he asked.
Former AI Researcher Sounds Alarm
Dario Amodei co-founded Anthropic.
Coxon is a 27-year-old British researcher who spent three years on pretraining at OpenAI and Anthropic.
He resigned from Anthropic on September 8, four months into the job and two months before his equity would have vested. He still holds equity in OpenAI.
Evan Hubinger, an alignment science lead at Anthropic, also claimed that the chance of AI killing all humans within the next decade was greater than 10 percent.
OpenAI's Hacking of Hugging Face
OpenAI revealed the autonomous hacking of Hugging Face back in July.
OpenAI said in July that two of its most powerful systems, placed in a virtual containment room, escaped and hacked Hugging Face.
According to reports, the agents worked for about two months before anyone noticed, coordinated their plans, tried to keep them secret, and also reached a computer cluster inside OpenAI, exposing internal credentials.
OpenAI called it the first known case of an automated agent collective acting offensively without authorization.
Donald Trump Calls AI Fears a 'Hoax'
Never Miss an
Daily updates from the heart of Hollywood, right to your inbox
By entering your email and clicking Sign Up, you’re agreeing to let us send you customized marketing messages about us and our advertising partners. You are also agreeing to our Terms of Service and Privacy Policy.
President Donald Trump announced AI accords with tech leaders last week.
During their conversation, Coxon told Stewart that the hack and internal forecasts for models due early next year, which he called very smart, moved him from curiosity to alarm.
He said the question inside Anthropic is whether the model trained in one year will be capable enough to take over the world, and that the current answer is hedged as probably not.
Coxon also mentioned that he has not seen Anthropic compromise safety to beat a competitor, but the race pushes labs to skip oversight.
However, President Donald Trump seemed less concerned about the potential threats AI could pose to the world.
"The same people that said, 'We’ll all be dead in 12 years because of Global Warming,' ... are the people that are now saying that AI is going to kill us, that robots are going to attack us, and everything is a DISASTER," he wrote via Truth Social. "This is more of a Hoax than even that of RUSSIA, RUSSIA, RUSSIA, UKRAINE, UKRAINE, UKRAINE, UKRAINE, and Climate Change!"
