Published Sep. 4, 2026at1:55pm
Last month, a man allegedly wrote in a chat with Anthropic’s Claude that he had purchased an AR-15 semiautomatic rifle and had CEO Dario Amodei “in his sights,” according to a police report obtained by The Standard. The company reported the chat to the San Francisco Police Department and said the person “was going to kill everyone at Anthropic and he purchased a firearm with the intent to kill the employees.”
Reached by phone, the man told The Standard he was “just fucking around” and was embarrassed about the ordeal. He declined to answer further questions. The Standard chose not to identify him because he has not been arrested or charged with a crime.
AI companies are taking seriously the rise in threats against their executives and workers. In April, OpenAI CEO Sam Altman’s Russian Hill house was targeted with a Molotov cocktail. Days later, security guards stopped a man (opens in new tab) in Anthropic’s lobby who said an executive would be killed, The Wall Street Journal reported. Public companies Palantir, Oracle, and Salesforce have all increased spending on executive protection in the last year.
The incident highlights the extent to which AI companies are on edge. It also raises questions about how they monitor usage of their tools at the individual level — including chats with bots like Claude that many assume are private — and the protocols they’ve developed for reporting troubling interactions with a chatbot to authorities.
“If internet companies don’t report it, they face significant liability if, in fact, a crime does occur,” said Eric Goldman, a law professor at Santa Clara University. “But as they become more hated, there’s more pressure on them to over-disclose knowing that some of the people they identify for law enforcement shouldn’t be targeted at all.”
He compared the system to the popular 2002 film “Minority Report,” in which police arrest potential murderers before they commit crimes.
Silicon Valley companies have traditionally complied with law enforcement requests about individual usage, and many older companies, including Google (opens in new tab) and Meta, (opens in new tab) have systems and policies in place for reviewing credible and imminent threats. It’s unclear whether the proactive reporting to police by Anthropic — and the categories of user actions that might prompt it — goes beyond what tech firms have typically done.
According to the police report, the individual made threats through Claude on Aug. 14, a Friday, at 2:30 p.m. The following Tuesday at 9 a.m., the company notified the SFPD, which sent officers to Anthropic’s 500 Howard St. location and determined that the person making the threats was not present. The employee who spoke with police said he was not immediately concerned about safety but wanted the event documented. The officer who wrote the report noted that the employee “refused to show me the messages due to Anthropic’s company policy.”
An Anthropic spokesperson told The Standard: “We banned this account, consistent with our standard practice, and referred the case to law enforcement. This is our safeguards process working as intended.”
The chatbot’s privacy policy allows (opens in new tab) the company to report any threats of violence within chats to law enforcement. An Anthropic spokesperson said its automated systems detect and block exchanges involving violence but declined to answer questions about whether it handles threats against employees differently from other mentions of guns, shootings, and targeted attacks.
The escalation of security precautions mirrors growing worries and protests about AI technology, which is quickly changing every aspect of the world. The companies have been criticized for their environmental footprint, their use of creatives’ work without consent, and the potential for their products to replace human workers. Big names in AI, including Amodei, Altman, and Bill Gates, have posted warnings about the impact of the technology, as has the pope.
That backlash has increasingly translated into real-world dangers, with threats against AI leaders and data centers up sevenfold (opens in new tab) from February to May, according to threat intelligence company Liferaft.
“Ultimately, the CEO is the person held responsible for decisions made by the company,” John Orloff, a security risk consultant, told The Standard. “People watching from the outside of the business put the blame squarely on the shoulders of the CEO.”
Anthropic is separately telling San Francisco-based employees to work from home due to a potential strike (opens in new tab) from the security workers — who are employed by a third party — at its Financial District building. The union and Allied Universal, the security company, have not agreed on terms for a new contract since the previous one expired in June.