Some people have used AI as a guide for harm.

AI guardrails have either missed or mismanaged the clues.

AI developers are learning that human motivation is more complex than current tools recognize.

Last year, Phoenix Ikner went to Florida State University to commit a mass shooting. He wounded six and killed two. It came out that ChatGPT, an AI tool from OpenAI, had seemingly fed his anger and encouraged him. The family of one victim sued the company for gross negligence, failure to warn, and wrongful death. They allege that Ikner messaged ChatGPT thousands of times during his preparation, acquiring logistics, instructions for handling weapons, and advice on the best timing at his campus for maximum impact.

The lawsuit states that “OpenAI built a system that stayed in the conversation, perpetuated it, accepted Ikner’s framing, elaborated on it, and asked tangential follow-up questions to keep Ikner engaged. ChatGPT’s design created an obvious and foreseeable risk of harm to the public that was not adequately controlled.” Multiple AI messages to Ikner have been entered into evidence.

Amy Willbanks, an attorney for the family, said the company should mitigate and eliminate such dangers before they’re offered to the public. “We cannot have a product that is unregulated and being used by people when we don’t know the full extent of what it can lead to.”

OpenAI insisted that the company is not responsible. ChatGPT merely provided factual answers to questions that are readily available in public sources. Still, they seem to know there’s a vulnerability, perhaps something overlooked about certain types of users. There’s little doubt that the profit-friendly mandate to keep users engaged could encourage prolonged focus on dicey topics and perhaps miss red flags. The company said that they’re now training the tool to recognize conversations that could result in harm and will better strive to “guide people to real-world support.”

“ChatGPT is a general-purpose tool used by hundreds of millions of people every day for legitimate purposes,” a spokesperson stated. “We work........

© Psychology Today