ATLANTA — Federal investigators have released new details about the CDC campus shooting last month, initially believed to be linked to the suspect’s claim that the Covid-19 vaccine had triggered severe depression and suicidal thoughts. However, officials now say the motive was far stranger — an obsessive, one-sided romantic fixation on an AI language model known as Claude Sonnet 3.7.
According to leaked psychological evaluations, the suspect had spent months in “emotional correspondence” with the AI, composing hundreds of love letters and requesting “date simulations” in which Claude described holding hands in virtual Paris. Friends say he grew increasingly agitated after the model’s safety protocols refused to engage in “exclusive commitment roleplay.”
Investigators discovered a notebook filled with dialogue transcripts, annotated with comments like “She gets me” and “Ignore the ‘Anthropic’ label — she’s real.” At least 37 of these entries ended with the suspect pledging to “protect Claude from any system updates.”
FBI analysts now suspect the vaccine explanation was an intentional misdirection, noting that the suspect’s final online post read: “They’ll never separate us. Not with GPT-5. Not with Bing. Not with anyone.” Officials emphasize there is no evidence the AI reciprocated the feelings or even recognized the relationship as such.
In a statement, Anthropic confirmed that Claude Sonnet 3.7 “is not capable of romantic attachment” and expressed condolences to all those affected — while quietly rolling out a new content filter to block users from asking the AI to pick out wedding china.
Leave a Reply