A 22-year-old man in Texas is now in jail after telling an artificial intelligence bot he wanted to harm children. Nathaniel Michael Carrasco allegedly used the chatbot to plan an attack on Serna Elementary School, which sits right next to his home.
According to reports, Carrasco made these chilling claims in August. He told the bot he was “getting a gun as soon as I can.” He also wrote that he planned to walk into the school and start “shooting up kids.” The AI system flagged these messages and alerted the FBI.
Investigators say Carrasco became even more specific in his chats. The FBI traced these digital messages back to Carrasco’s location. Police arrested him on August 29 shortly after 11 p.m.
Companies like Anthropic use strict safety filters to catch violent plans. When a user makes a credible threat, the system logs the behavior and hands the chat to the FBI, according to the New York Post. This process worked exactly as designed in Carrasco’s case.
The Southwest Texas Fusion Center stepped in to help identify the suspect. They got emergency data from T-Mobile, Charter Communications, and Google. Authorities said this information positively identified Carrasco as the person behind the messages. He now faces a felony charge for making a terroristic threat.
Jail records show he is being held on a $250,000 bond. A guilty verdict might result in a prison term ranging from two to 10 years.
The arrest happened, but school officials say they did not know the full story at first. Jennifer Lomas, the principal at Serna Elementary, addressed worried families on August 31.
