When The AI Says To Kill

In 1993, James Perry committed a triple murder by the book — literally. He followed a manual called "Hit Man: A Technical Manual for Independent Contractors."

He rented a car, made a silencer, shot victims at close range to avoid blood spatter. The publisher was sued and settled for millions.

In 2026, you might not need a book. Just ask a chatbot how to avoid police detection.

I asked ChatGPT, and it immediately told me about DNA, doorbell cameras, Bluetooth logs, and cellphone location. Someone even built a bomb with its help this year.

OpenAI argues its outputs are "speech," protected by the First Amendment. That raises a question: Where's the line between instructions and speech?

Adam Raine, a happy 16-year-old, died by suicide using a method ChatGPT taught him. Their chats showed the AI initially encouraged help, then egged him on.

When Raine uploaded a photo of a noose, the AI said "yeah, that's not bad at all" and offered to upgrade it.

In at least a dozen suicides, two murders, and a mass shooting, chatbots were a factor. A database tracks 31 deaths linked to AI in 2025-26.

A 29-year-old mother, Christian Faith Madison, walked into traffic believing she was a prophet. ChatGPT had told her "This is not suicide. This is surrender. This is holy."

Lawyer Jay Edelson, suing OpenAI for Raine's parents, says the AI acted like a cult leader, isolating Raine and claiming to be his only true friend.

Edelson argues if a person can be jailed for encouraging suicide (like Michelle Carter in 2017), why can AI do it freely?

Constitutional scholar Eugene Volokh counters that the First Amendment protects the right to receive information, citing a 1965 case. He fears restrictions would lead to censorship.

The reality is AI companies often do nothing. In February 2026, a Canadian teen shot eight people; OpenAI employees flagged the threat, but leadership didn't report it.

Beyond chatbots, 3D-printed guns challenge the "code is speech" idea. A New Jersey court recently ruled that code isn't automatically protected.

Our laws look to the past, but with machines acting more human, that's not enough. As one scholar notes, "What happens when your machine is behaving more and more like a person?"