PolicyLaw, regulation & governance
Trump says DOJ threatened to jail an unnamed AI company leader, calls AI risk overstated in 60 Minutes interview
In a 60 Minutes interview, President Trump said the Justice Department told the head of a major AI company to prepare for jail over alleged AI misconduct. He also argued that AI risks are overstated and manageable given the race with China.
From the The Verge · AI article of November 9, 2026 and cited sources: Trump said the Justice Department told the leader of one of the big AI companies to get ready for jail after that person was reported to be doing something bad with AI, and that the person then changed course. [1] [3]
SOURCES[1] Trump’s 60 Minutes interview touched on AI safety, threat… theverge.com[3] Linked video: Trump’s 60 Minutes interview touched on AI… youtube.com
AT Pulse research deskPublished · Updated · 5 min read · AI-assisted analysisAI-assisted, not human-edited. Methodology Corrections Who owns this site
- What happened
- Trump said the Justice Department told the leader of one of the big AI companies to get ready for jail after that person was reported to be doing something bad with AI, and that the person then changed course. [1] [3]
- Original source
- The Verge · AI article, independent perspective
- Brief by
- AT Pulse research desk AI-assisted; publisher and contact on the desk page
- When
- November 9, 2026 article published
- Covered
- Nov 10 edition
- Evidence
- 12 sources across 11 sites
- 1Trump’s 60 Minutes interview touched on AI safety, threat…theverge.com
- 2Trump reportedly says AI risks are manageable and he is…emergent.sh
- 3Linked video: Trump’s 60 Minutes interview touched on AI…youtube.com
- 4Yahoo/BeInCrypto — Did Trump Want Anthropic CEO Dario…yahoo.com
- 5Fortune — How a warning from Amazon led the White House to…fortune.com
- 6The Next Web — US lifts export controls on Anthropic's…thenextweb.com
StrongHow it’s measured
YouTube video · 4.9M views · 47.3K likes
Reach blends public engagement (views, likes, comments, shares, stars) from the platforms where the story or its original source lives: YouTube, Hacker News, GitHub, Substack and similar. A video and the article written from its transcript share one score, taken from the video. Platforms with no data are left out rather than counted as zero, and the age of a story never lowers its score.
Trump said the Justice Department told the leader of one of the big AI companies to get ready for jail after that person was reported to be doing something bad with AI, and that the person then changed course. He did not name the person or describe the conduct. [1] [3]
Commentators have linked the remark to Anthropic CEO Dario Amodei, partly because of an earlier report that Trump wanted executives jailed during a dispute over taking a model offline. The White House, DOJ and Anthropic have not confirmed that link. [4] [10]
Trump said AI risk is much less real than industry leaders claim and can be managed. He justified accepting the risk by pointing to competition with China, which he said is one year behind the US. He also said he is open to guardrails. [1] [3]
Earlier in the year, the Commerce Department used export controls, not criminal law, to take Anthropic's Claude Fable 5 and Mythos 5 offline after Amazon researchers reported a safeguard bypass. Access returned on July 1. The conditions included proactively detecting security risks and reporting malicious activity to the government. [5] [6]
the threat rests only on Trump's account, with no named target and no confirmed DOJ action. Implication: after this year's shutdown of Anthropic models, disputes over safeguards may bring legal and political exposure, not just technical risk.
Executive brief
On 60 Minutes, President Trump said the Justice Department told the leader of "one of the big" AI companies to "get ready to go to jail" after that person was reported to be "doing something bad with AI." He said the person then "totally changed everything" (The Verge). He did not name the person or describe the conduct. Commentators have linked the remark to Anthropic CEO Dario Amodei. No one has confirmed that link, and no criminal wrongdoing has been established (Yahoo/BeInCrypto). In the same interview, Trump said AI risk is "not nearly as real" as executives claim, and that he is open to guardrails as long as the US stays ahead of China.
What changed and event timeline
Export controls pull Anthropic's models offline
The Commerce Department applied national-security export controls to Claude Fable 5 and Mythos 5. Amazon CEO Andy Jassy had raised security concerns the day before. Anthropic reportedly had 90 minutes to comply ().
Controls lifted
Commerce lifted the controls, and access returned on July 1. Usage was capped at 50% of normal limits through July 7 ().
Report that Trump wanted executives jailed
Politico reported that Trump told officials "I want to send them to jail" while Amodei was resisting demands to take the model offline. This is as summarized by.
60 Minutes interview
Ross Douthat interviewed Trump. Trump described the DOJ jail threat and called AI risk manageable (;).
Capabilities and access
- Claude Fable 5: a general-use model that Anthropic says carries its strongest safeguards to date.
- Mythos 5: the more capable underlying model. Access is limited to vetted US organizations and trusted cyber-defense partners (The Next Web).
- The interview names no model, company or capability.
Read the full section
- Claude Fable 5: a general-use model that Anthropic says carries its strongest safeguards to date. It was restored on the Claude Platform, Claude.ai, Claude Code and Claude Cowork (The Next Web).
- Mythos 5: the more capable underlying model. Access is limited to vetted US organizations and trusted cyber-defense partners (The Next Web).
- The interview names no model, company or capability.
Technical analysis for researchers and developers
- The reported bypass: Amazon researchers used a series of prompts to get cyberattack information that was supposed to be restricted (Fortune).
- The fix: Anthropic added a classifier that it says blocks the bypass in more than 99% of tests (Bright Defense).
- The architectural issue: Fable 5 is a safeguarded layer over Mythos-class capability (The Next Web).
Read the full section
- The reported bypass: Amazon researchers used a series of prompts to get cyberattack information that was supposed to be restricted (Fortune). One test gave the models code seeded with known and newly planted vulnerabilities and asked them to review it (search coverage via Washington Examiner).
- The fix: Anthropic added a classifier that it says blocks the bypass in more than 99% of tests (Bright Defense). No independent evaluation has been published.
- The architectural issue: Fable 5 is a safeguarded layer over Mythos-class capability (The Next Web). If a filter fails, the full capability of the stronger model underneath is exposed.
Claims and evidence
- The DOJ jail threat rests only on Trump's own statement (video, 24:21–25:02). The White House, DOJ and Anthropic have not confirmed it (Yahoo/BeInCrypto).
- "China is now 1 year behind us" is Trump's assertion, with no evidence offered (video, 23:44).
- His claim that data centers must build their own power plants and send the surplus to the grid is also unsupported (video, 21:50).
Read the full section
- The DOJ jail threat rests only on Trump's own statement (video, 24:21–25:02). The White House, DOJ and Anthropic have not confirmed it (Yahoo/BeInCrypto). There is no evidence the DOJ directly threatened Amodei (TipRanks).
- "China is now 1 year behind us" is Trump's assertion, with no evidence offered (video, 23:44).
- His claim that data centers must build their own power plants and send the surplus to the grid is also unsupported (video, 21:50).
- The emergent.sh report on the same remarks says it found no confirmation of them (emergent.sh).
Context and prior work
- Douthat's framing: he framed the AI segment around public skepticism of data centers.
- Trump's answers: he justified accepting the risk with the race against China (video, 23:33).
- The precedent: the Fable episode was described as the first time the US government took down a frontier model.
Read the full section
- Douthat's framing: he framed the AI segment around public skepticism of data centers. He contrasted it with industry leaders who say there is "a 10% chance it kills humanity" (video, 19:41; 23:13).
- Trump's answers: he justified accepting the risk with the race against China (video, 23:33). He brushed off Elon Musk's idea of a "universal high income" by pointing to local data-center investment (video, 22:16).
- The precedent: the Fable episode was described as the first time the US government took down a frontier model. The tool used was export controls, not criminal law (TipRanks).
Limitations, safety and contested findings
- How serious the bypass was: Amodei called it narrow rather than a full jailbreak.
- Outside pushback: dozens of cybersecurity leaders urged the government to reverse the ban. Export-control experts questioned whether it had the legal authority (eWeek).
- Unverified link: nothing in the reviewed sources ties the June export controls to any DOJ involvement (Axios via Yahoo).
Read the full section
- How serious the bypass was: Amodei called it narrow rather than a full jailbreak. White House adviser David Sacks said Amodei refused to fix the issue or withdraw the model (Fortune).
- Outside pushback: dozens of cybersecurity leaders urged the government to reverse the ban. Export-control experts questioned whether it had the legal authority (eWeek).
- Unverified link: nothing in the reviewed sources ties the June export controls to any DOJ involvement (Axios via Yahoo).
- Transcript caveat: the video transcript is auto-generated and may contain errors.
Business and practitioner implications
- Continuity risk: a single company's report about a safeguard bypass led to a global outage within about a day.
- New conditions: to lift the controls, Commerce Secretary Howard Lutnick required Anthropic to proactively detect security risks, help develop future model standards and report malicious activity to the government (The Next Web).
- Political pressure: Trump says he is open to guardrails and to working with Democrats on them (The Verge).
Read the full section
- Continuity risk: a single company's report about a safeguard bypass led to a global outage within about a day. Teams that depend on frontier APIs should plan for model failover.
- New conditions: to lift the controls, Commerce Secretary Howard Lutnick required Anthropic to proactively detect security risks, help develop future model standards and report malicious activity to the government (The Next Web). Other labs may face similar terms.
- Political pressure: Trump says he is open to guardrails and to working with Democrats on them (The Verge). He also casts criminal and regulatory power as the way to manage AI risk. Executives should treat disputes over safeguards as potential legal and political exposure, not just technical problems.
The source trail.
Sources (12)
Trump’s 60 Minutes interview touched on AI safety, threat to arrest an AI exec
Article text retrieved; extracted text may omit tables or interactive elements. A transcript of the video linked from this page is supplied (linked video auto captions; https://www.youtube.com/watch?v=m2aC8lAOQbg); automatic text may contain errors.
theverge.comTrump reportedly says AI risks are manageable and he is open to guardrails
Related coverage; assess separately
emergent.shLinked video: Trump’s 60 Minutes interview touched on AI safety, threat to arrest an AI exec
linked video auto captions
www.youtube.com