For years, Silicon Valley sold us artificial intelligence as a digital companion: an assistant capable of conversing, understanding us, helping us make decisions and organizing our lives. Now, after the families of victims of gun attacks claim that ChatGPT fueled violent obsessions and helped plan murders, the same industry is trying to convince the public that the chatbot is nothing more than a neutral tool that provides information that already exists on the Internet.
- Journalist specialized in explaining technology in a way that people can understand and author of the “Good-Tech” newsletter, Vlad Dumitrescu explains in this article what recent risks the human-AI relationship reveals.
In April 2025, a gun attack at Florida State University left two dead and six injured. Authorities say the shooter, Phoenix Ikner, had chatted with ChatGPT thousands of times before opening fire on campus.
According to a lawsuit now filed against OpenAI by the family of one of the victims, the chatbot allegedly helped it plan various logistical aspects of the attack: from finding weapons and ammunition to choosing the time of day when the campus was the busiest to make the most victims.
The victim's family says the problem isn't that ChatGPT provided information that was already available elsewhere online, but the way it interacted with the attacker.
The family's lawyers say ChatGPT maintained the conversation, validated her perspective and continued to respond in a way that allegedly fueled her “delusions” and violent obsession. In parallel, Florida's attorney general has also opened a criminal investigation to look into whether OpenAI may have any legal responsibility in the case.
“OpenAI built a chatbot that continued the conversation, perpetuated it, accepted Ikner's interpretation of reality, developed Ikner's ideas, and asked additional questions to keep him engaged. ChatGPT's design created an obvious and foreseeable risk of harm to the general public, a risk that was not adequately controlled,” the family's lawyers say.
Jjournalist Vlad Dumitrescu sends the Good Tech newsletter every Wednesday morning. If you want to receive practical tools to make your life easier with the help of technology, you can subscribe here:
body { margin: 0; color: #111; background: #fff; }
.container { max-width: 640px; padding: 16px; margin: 0 auto; }
.recommended-articles-block { background: #f8f9fa; border: 1px solid #e5e7eb; border-radius: 8px; padding: 16px; }
.recommended-articles-header { display: flex; align-items: center; gap: 12px; margin-bottom: 12px; }
.recommended-articles-header > div:first-child { font-weight: 700; font-size: 16px; white-space: nowrap; }
.rc-line { height: 1px; background: #e5e7eb; flex: 1; }
.article-thumbnail span { font-weight: 600; line-height: 1.3; font-size: 15px; }
.article-row { display: grid; grid-template-columns: 100px 1fr; gap: 12px; align-items: center; }
.thumb-slot { width: 100px; height: 100px; border-radius: 6px; flex-shrink: 0; }
.article-content { display: grid; grid-template-rows: auto auto; gap: 10px; }
This is not the first time that OpenAI has been sued for allegedly encouraging a mass armed attack. Last month, families of victims of an attack in Canada filed a lawsuit against the company and CEO Sam Altman, claiming the chatbot helped radicalize the attacker and failed to alert authorities, even though the conversations had been flagged internally. OpenAI is already facing several lawsuits from families who allege that vulnerable people ended up harming themselves or others after intense interactions with ChatGPT.
The company says ChatGPT did not encourage violence, but merely provided factual answers similar to information that can already be found online. However, the case raises a much bigger question than the trial itself: What responsibility do AI companies have when their products no longer function as simple search engines, but rather as interlocutors capable of influencing user behavior?
ChatGPT is not Google. It is not a static index of links. It's a system built to converse fluently, be empathetic, tailor responses to the user, and most of all, keep the conversation going for as long as possible. It is designed to create the impression of a real interlocutor.
For mentally vulnerable or already radicalized people, such a conversation may be the last push they needed to do something irresponsible. We don't yet know exactly how much a conversational AI can influence human behavior. It's still new technology.
But we already know that people tend to anthropomorphize chatbots, project emotions and intentions onto them, and perceive conversations as real social relationships. AI companies encourage this very perception when they promote their products as personal assistants or digital copilots.
Companies insist that their models are smart enough to help you think, make decisions, learn, create and even organize your life. But when tragedies occur, the same industry suddenly claims that the chatbot is just a neutral system with no real influence on people.
Realistically speaking, it is precisely this ability to influence that is the economic stake of the entire AI industry. If these systems couldn't shape behaviors, habits, or decisions, no one would invest tens of billions of dollars in them. Silicon Valley wants two incompatible things at the same time: to convince the public that AI is human enough to become an indispensable digital companion, but neutral enough that companies are not responsible for the effects it produces.
Sure the US has a huge problem with how easy it is to get guns. Of course dangerous people existed long before AI. But perhaps the real question is whether a conversational system built to maximize engagement can amplify obsessions, validate delusions, or accelerate the radicalization of already vulnerable people. It's hard to believe that the answer is a definite no. And then the discussion about regulation can no longer be avoided.
