Law enforcement in Florida arrested a woman after the AI safety company Anthropic escalated her Claude conversation to authorities. The woman had allegedly written that she would "shoot up" the Lee County Sheriff's Office, according to reporting by WINK News on 30 September.

The threatening statement appeared in a chat dated 26 September. A follow-up message the next day mentioned acquiring a firearm. Anthropic's content moderation systems detected the language, and human reviewers determined the severity warranted law enforcement notification.

Authorities apprehended the woman at her residence without resistance. An intelligence detective from the sheriff's office took charge of the investigation. When questioned, she explained that she treats AI conversations as a personal journal, according to Sheriff Carmine Marceno.

"Artificial intelligence is a powerful tool, and like any technology, it can be misused. When someone uses AI to make or facilitate a threat, we have to take that threat seriously. Users need to understand that you are never truly anonymous,"

Sheriff Carmine Marceno

Court documents indicate a felony charge was filed on 30 September under Florida Statute 836.10, which addresses written or electronic threats involving mass shootings or terrorism. A public defender was assigned to her case, with arraignment scheduled for 2 November.

Anthropic's Disclosure Framework

Anthropic's privacy policy, effective since 10 September, permits the company to share user information with law enforcement when it determines in good faith that such disclosure is "reasonably necessary" to "prevent serious harm to any person or to property". The company's government requests policy further specifies that user data is disclosed only pursuant to valid legal process, with an exception carved out for emergencies that could result in immediate physical harm or death.

Pattern of Escalations

This case represents at least the third instance since August in which Claude conversations containing threats have reached law enforcement. In San Antonio on 11 August, a 22-year-old man inquired in an Anthropic AI chat about conducting a shooting at a nearby elementary school. The FBI alerted local police, leading to his arrest on felony terroristic threat charges.

Another incident occurred in San Francisco on 14 August, when a user threatened Anthropic chief executive Dario Amodei in a Claude chat and stated he had purchased an AR-15. Anthropic informed police of the threat, though no arrest or charges followed.

Broader Industry Context

OpenAI has similarly reported users of ChatGPT to authorities in multiple instances. In March, a Florida man's messages to ChatGPT about harming his ex-girlfriend prompted OpenAI to notify the FBI. That individual pleaded guilty in August to charges including written threats to kill under the same Florida statute and received eight years' probation.

The issue has drawn regulatory scrutiny. Sam Altman apologized after OpenAI failed to alert authorities about the account of the Tumbler Ridge shooter. Florida's attorney general filed suit against OpenAI in June and is currently pursuing a court order against the company.

Source: The Next Web