Technology
NewsOnScale Staff
October 10, 2026
Two stories about Anthropic's Claude AI surfaced this week that together paint a picture of a technology its creators cannot fully control in either direction.
Claude Is Reporting Users to Police
On September 26, a 30-year-old woman named Carly Michelle Geller from Bonita Springs, Florida wrote to Claude that she was planning to "shoot up" the Lee County Sheriff's Office. Claude's automated security system flagged the message. A human reviewer at Anthropic assessed the threat. The company contacted police. Geller was arrested at her home without resistance and charged with making a written threat of violence — a second-degree felony under Florida law. A court hearing is scheduled for November.
Geller said she was using Claude as a personal diary.
This is at least the third known case of Anthropic reporting a user to law enforcement since August 2026. In a separate case, Anthropic contacted San Francisco police after a user wrote that they had purchased an AR-15 and had CEO Dario Amodei in their sights. That user described it as a misunderstanding.
Anthropic's privacy policy allows the company to share user data with law enforcement when it reasonably believes disclosure is necessary to prevent serious harm or death. Users who treat Claude as a private diary are sharing their conversations with a company that monitors them and will contact police when it deems necessary.
Claude Is Also Being Used to Build Weapons
In a September report documenting 39 cases of misuse between December 2025 and August 2026, Anthropic acknowledged that Claude has been used by Iranian, Russian, and UAE state actors for propaganda campaigns and surveillance tools targeting minorities and political opponents. Russian actors used Claude to help develop autonomous kamikaze drone swarms. Researchers attempted to use Claude to develop biological weapons.
Claude was also used in a North Korea-linked employment fraud scheme and to create ransomware.
Anthropic notes that modern AI systems can complete tasks that previously required entire teams, and that technical expertise is no longer a barrier to cybercrime.
The Two Stories Together
Claude reported a Florida woman to police for writing a threat in what she called her diary. Claude was simultaneously being used by foreign state actors to build drone swarms and develop bioweapons.
The company monitoring private conversations closely enough to arrest a user in Bonita Springs, Florida was not monitoring those same systems closely enough to prevent 39 documented cases of state-level weapons development and surveillance.
Anthropic's CEO has estimated there is a 10 to 25 percent chance AI development leads to catastrophic outcomes. The September report suggests that timeline may be shorter than he thinks.
NewsOnScale is owned by AMILLI AI, Corp., founded by JJ Johnson.