Man arrested after AI chatbot flagged his plan to shoot children at nearby school
A San Antonio resident who lives next to Serna Elementary used an AI chatbot to detail a plan to obtain a gun and attack the school, prompting an FBI alert and his arrest.
Nathaniel Michael Carrasco, a San Antonio man residing adjacent to Serna Elementary, engaged Anthropic's Claude chatbot with violent intentions, asking it to locate graphic images of deceased schoolchildren and stating he would acquire a firearm and enter the school to shoot kids. Anthropic's safety filters recorded the exchanges and alerted the FBI's National Threat Operations Center, which warned local police on Aug. 28.
Investigators used emergency disclosures from T-Mobile, Charter Communications and Google to confirm Carrasco's identity. He was detained by San Antonio police on Aug. 29, charged with making a terroristic threat, and placed in Bexar County jail with a $250,000 bond, facing a potential sentence of two to ten years. The school principal, Jennifer Lomas, later disclosed that the institution had only received partial information and only learned of the specific threat after media reports emerged. This case highlights how AI monitoring tools can trigger law-enforcement action on online threats.
Why it matters
It shows AI safety systems can help prevent potential school shootings by alerting authorities to credible threats.
In this story
