Ask the president how he plans to keep AI safe, and you might expect an answer about laws, agencies or testing rules. Donald Trump told a story about a phone call instead. In his interview with Ross Douthat on 60 Minutes, which aired Sunday, he said his administration once warned the leader of a major AI company to “get ready to go to jail.” As UNILAD Tech reported, Trump says the threat worked and the executive “totally changed everything.”
He wouldn’t say who it was, which company, or what they were supposed to have done. That hasn’t stopped the guessing. Here’s exactly what Trump said, the context around it, and why the way he described it matters more than the mystery of the name.
The short version
- Trump said someone at an AI company “was reported to us to be doing something bad with AI,” and the government told them they’d be in jail the next morning.
- Asked if it was the leader of a major AI company, he said: “One of the big ones.”
- He said he didn’t think it was done “maliciously,” but it “coulda been.”
- No name, no company, no date, and no description of the conduct. The White House and DOJ didn’t immediately comment.
- Some outlets have speculated about Anthropic because of its public fight with the administration. Nothing confirms that.
What Trump said, word for word
The exchange came during a longer conversation about whether the US should slow down advanced AI. Douthat pointed out that some AI leaders openly say their technology carries a real chance of catastrophe. Trump said the risk is “not nearly as real” as they claim and “can be managed.” Then he explained by whom.
“The DOJ. We had a situation where somebody was reported to us to be doing something bad with AI. We called them very strongly and said, ‘Get ready to go to jail. Please give your wife a kiss good night.’ … I can’t tell you. But we said, ‘Do you like the color orange?’ … ‘Go home right now. See your wife and children. Give ’em a kiss good night,’ … ‘because tomorrow morning you’re goin’ to jail.’ Let me tell you, this person totally changed everything. Totally. ‘Yes, sir. Yes, sir.'”
Douthat asked: “Was it a leader of a major AI company?” Trump answered: “One of the big ones. Yeah.” He added that he didn’t think it was “done maliciously. But it coulda been malicious.”
The “color orange” line is a reference to prison jumpsuits. The overall picture is of a warning delivered by phone, outside any court, and a CEO who backed down.
So who was it?
We don’t know, and neither does anyone else outside the administration. Trump’s own phrase, “one of the big ones,” narrows it down only a little. Over the past two years he has met with nearly every major AI boss and collected public commitments from most of them.
Some outlets have pointed at Anthropic and its CEO, Dario Amodei. The reasoning is circumstantial. Anthropic has been in an open fight with the administration over limits it puts on military use of its Claude models, and it has sued the government over what it calls an “unlawful campaign of retaliation.” A White House spokesperson previously said Trump can’t let a “radical left, woke company” put national security at risk. There were also reports earlier this year that Trump wanted Anthropic executives jailed.
None of that confirms Anthropic was the company in Trump’s story, and the June reporting never established that anyone there received a criminal threat. Trump’s description also doesn’t fit neatly: he described a person who was caught “doing something bad” and then fell into line, not a company that’s still suing him. It could be any of the firms in the chart above, or one that isn’t in it. Until someone on the record names names, treat every guess as a guess.
Why the way he said it matters
The more interesting part of the answer isn’t the who. It’s the how. Asked how AI risk is managed, the president didn’t point to a law or a regulator. He pointed to the threat of prosecution, delivered informally, by phone.
That’s notable because there is still no broad federal AI safety law in the US. Companies building frontier models operate under voluntary commitments and an executive order, not statutes written for AI. Prosecutors can only charge someone under existing laws, such as fraud, export control or computer crime statutes. Trump didn’t say which law, if any, this person was thought to have broken, or whether anything was ever filed.
| How AI is policed today | Binding? | What it covers |
|---|---|---|
| Federal AI safety law | Doesn’t exist yet | Bills in Congress, none passed |
| Sept 2026 White House safety accord | Voluntary (“morally binding,” per Trump) | Internal controls, outside testing, board oversight |
| Executive orders | Bind federal agencies, not directly companies | Federal policy, procurement and priorities |
| Existing criminal law | Yes | Fraud, export controls, hacking and other crimes, whether or not AI is involved |
| An informal threat of jail | Not a legal process | Whatever the person on the other end of the phone agrees to |
That gap is exactly what critics worry about. If the government believed a crime was committed, the normal path is an investigation and, if the evidence holds up, charges. If it wasn’t a crime, a jail threat looks like pressure without a legal basis. Either way, the public is left without the most basic facts. We looked at how the administration has tried to shape AI policy by decree in our breakdown of Trump’s “Super Intelligence” executive order, which leans on agencies rather than new law.
Guardrails, but not too many
Right after the jail story, Douthat asked whether Trump would work with Democrats on AI guardrails. His answer was a qualified yes. “I don’t mind guardrails. But I don’t wanna put us in a deep second position. We’re leading number one. And China has no guardrails, by the way,” he said. Douthat countered that China has “the guardrails of being a totalitarian state.” Trump agreed that in China, “whatever President Xi says goes,” adding: “With me, it’s pretty strong, but it’s not quite that strong.”
He also claimed China is now “one year behind us” and said the government has “very brilliant people watching every one of these companies.”
The question of guardrails is about to get real. Democrats won both chambers of Congress last week, partly by campaigning on electricity bills and data centers. We covered what a Democratic Congress can actually do about AI data centers, and AI safety legislation is likely next on the list. Geoffrey Hinton has already warned lawmakers that the current crop of bills will die in January if nothing moves, as we reported in his one-year warning to Congress.
The questions nobody has answered
Trump’s anecdote leaves a lot hanging, and the White House and Justice Department didn’t immediately respond to requests for comment from Business Insider. These are the things worth watching for:
- What was the “something bad”? A safety failure, a data issue, a national security concern and a business dispute would all mean very different things.
- Was there a real case? If prosecutors believed a crime happened, was it dropped because the person “changed everything”? That’s not how criminal law is supposed to work.
- Who made the call? Trump credited “the DOJ.” Whether it came from career prosecutors or political staff matters a lot.
- Will Congress ask? With Democrats taking over committees in January, this is the kind of remark that tends to turn into a letter, then a hearing.
Sources and further reading
- UNILAD Tech: Trump claims major AI boss was threatened with jail and it made them change everything
- CBS News via Yahoo: Full transcript of Trump on 60 Minutes
- Business Insider: Trump says the administration once threatened the leader of a big AI company with jail time
- Benzinga: “Get ready to go to jail”: Trump reveals how DOJ dealt with alleged AI misconduct
About this article: GeekBlog covers U.S. technology news, AI, phones, smartwatches and gaming. Every story is written and checked under our Editorial Policy. Spotted a mistake or have a story tip? Contact our editors.
