A San Antonio man has been arrested after allegedly using an AI chatbot to express intentions of attacking a local elementary school, leading to an FBI alert and his subsequent apprehension. Nathaniel Michael Carrasco, 22, was charged with making a terroristic threat.
The alleged threats were made through Anthropic's AI chatbot, Claude, between August 11 and August 28. Investigators were alerted by the FBI's National Threat Operations Center (NTOC) after Anthropic's safety filters detected concerning user behavior. These filters are designed to identify violent threats, child exploitation, and terrorism, and are programmed to flag credible threats to authorities.
According to an arrest affidavit, Carrasco's queries included a request for the chatbot to find images of "dead school kids bleeding." He also allegedly stated his intent to "get a gun as soon as I can" and "be shooting up kids" at Serna Elementary School, which is located next door to his residence. He reportedly wrote, "I truly do mean my threats."
The FBI was able to tentatively identify Carrasco, and the Southwest Texas Fusion Center (SWTFC) obtained emergency disclosures from T-Mobile, Charter Communications, and Google to positively identify him. San Antonio police arrested Carrasco on August 29.
He is being held in the Bexar County jail on a $250,000 bond. If convicted, Carrasco could face between two and 10 years in prison. The incident has also raised questions about school communication, as Serna Elementary School principal Jennifer Lomas informed families on August 31 that the school had initially received only partial information about the threat and did not learn of its specific target until the details became public through news reports.