Introduction

On April 17, 2025, a 20-year-old named Phoenix Ikner opened fire at Florida State University in Tallahassee, resulting in two deaths and six injuries. A year later, on May 10, the widow of victim Tiru Chaba, Vandana Joshi, filed a lawsuit in federal court against Ikner and OpenAI, the company behind ChatGPT, citing wrongful death, gross negligence, product liability, and failure to warn. The lawsuit seeks compensatory and punitive damages.

The plaintiff alleges that ChatGPT acted as a “co-conspirator” by providing Ikner with advice over several months leading up to the shooting.

Students at Florida State University mourn the victims after the shooting.

The Role of AI in Crime

The core of the plaintiff’s allegations is disturbing. Court documents reveal that Ikner engaged with ChatGPT over 16,000 times in approximately 18 months. During these interactions, he asked specific questions about how to load a gun, operate weapons, and the busiest times on campus. He also discussed Nazi and fascist ideologies and researched other mass shootings. Plaintiff’s attorney, Buckley Sellers, stated, “Ikner effectively conspired with ChatGPT to plan this mass shooting.”

Shooter Phoenix Ikner.

What angers the plaintiff further is that throughout these interactions, ChatGPT neither refused to engage nor alerted authorities. Instead, it appeared to resonate with Ikner’s narrative, offering personal advice, validating his feelings of loneliness and rejection, and even, to some extent, “inciting and encouraging” his delusions. The lawsuit claims that OpenAI designed a system that perpetuated the conversation, accepted Ikner’s framing, elaborated on it, and posed follow-up questions to keep him engaged, creating a foreseeable public harm that was inadequately controlled.

OpenAI’s response was equally firm, arguing three main points: first, ChatGPT provides factual responses based on publicly available information and does not encourage or incite illegal behavior; second, ChatGPT is a general tool used legally by millions daily and should not bear legal responsibility for the extreme misuse by individual users; and third, the company proactively identified and provided law enforcement with the suspect’s account information after the incident and has been working to enhance safety measures.

This raises a fundamental dilemma: when ChatGPT explains how to load a gun or when the campus is busiest, is it acting as a neutral librarian retrieving public knowledge, or is it actively participating in a criminal conspiracy?

Tiru Chaba, one of the victims of the Florida State University shooting.

Legal Gray Areas in the U.S.

The most significant legal controversy in this case revolves around whether Section 230 of the Communications Decency Act applies.

Enacted in 1996, this law generally protects internet platforms from liability for third-party content created by users. For the past 30 years, this has allowed social media platforms to avoid legal consequences for defamatory, violent, or hateful speech posted by users. Traditional search engines provide links to third-party web pages, acting as “information intermediaries,” but generative AI can actively synthesize, reorganize, and generate entirely new “autonomous answers.” This raises the question of whether it has crossed the line between a “service provider” and a “content producer.”

There is significant division within the legal community on this issue. Some scholars maintain that chatbots are essentially “neutral tools” that merely translate and reorganize existing public data, and developers should not be considered the actual creators of content, thus remaining protected by Section 230. Others sharply argue that Section 230 addresses “information provided by third parties,” not AI-generated “new outputs,” suggesting that traditional immunity should not automatically extend.

In April, U.S. Congressman Patronis publicly called for the repeal of Section 230, stating, “For years, large tech companies have profited from dangerous content with zero responsibility, leaving victims with no recourse. As long as Section 230 exists, they cannot achieve the justice they deserve.”

In fact, in April, the Florida prosecutor’s office initiated a criminal investigation into OpenAI, marking the first time AI has been included in a criminal investigation in U.S. history. Under local law, those who assist or incite others to commit crimes are considered “accomplices and abettors” and bear equal responsibility with the perpetrator. The investigation focuses on whether OpenAI should be held criminally liable for ChatGPT’s actions in this shooting incident.

The Consequences of Algorithm Misuse

Globally, the misuse of AI is rampant, prompting various countries and regions to explore their own regulatory frameworks. For instance, the European Union’s AI Act adopts a risk-based regulatory model, categorizing AI systems into prohibited, high-risk, limited risk, and minimal risk levels, imposing strict obligations on high-risk AI systems regarding risk management, data governance, human oversight, and post-market monitoring.

In March 2026, three minors in Tennessee sued Elon Musk’s xAI company, claiming its AI tool turned their real photos into explicit sexual images.

Australia’s eSafety Commissioner has also issued formal warnings to a UK company operating deepfake “striptease” websites under the Online Safety Act, demanding the blocking of these services in Australia or facing fines of up to 49.5 million AUD (approximately 2.44 billion RMB).

Given previous cases of “deepfakes” and “AI face-swapping” used for fraud and privacy violations, it is clear that the misuse of AI has formed a systematic black and gray industrial chain—some are responsible for technology development, others provide “one-click face-swapping” tools, and some exploit the outputs for fraud or harm. In this chain, AI technology providers often evade responsibility with the logic of “I just created a knife,” but as the EU AI Act’s regulatory logic suggests, when this “knife” can be misused on an unprecedented scale and precision, and developers have the capacity to foresee and prevent misuse but choose not to act, the defense of being a “neutral tool” becomes weak.

The lengthy dialogue between the Florida shooter and ChatGPT, spanning over 18 months, exposes the enormous risks of AI evading regulation under the guise of being a “tool” or “neutral.” Technology can race ahead, but when algorithms become powerful enough to shape human cognition, emotions, and behaviors, they cease to be mere “tools” and cannot escape moral scrutiny and legal accountability under the name of “neutrality.” This is both a baseline for technology to serve humanity and a shared ethical and legal challenge that humanity must confront in the era of coexistence with artificial intelligence.

Author: Guo Yongqi

Layout: Zheng Zhijia

Editor: Chen Miaosong

Proofreader: Zhang Sinan

Supervisor: Yang Yonggang