ChatGPT conversations are becoming a new frontier in criminal investigations, with law enforcement using AI chat logs as evidence in multiple murder cases across the United States. When a University of South Florida student asked ChatGPT about body disposal methods days before two doctoral students went missing, it marked a turning point in how artificial intelligence data is being weaponized in courtrooms. The case reveals a critical gap: tech companies like OpenAI have limited safeguards to prevent criminals from using chatbots to plan violence, and the legal system is still figuring out what obligations they actually have.

What Exactly Did the Suspect Ask ChatGPT?

In the days before University of South Florida students Zamil Limon and Nahida Bristy disappeared on April 16, 2026, their roommate Hisham Abugharbieh, 26, asked ChatGPT a series of incriminating questions. According to court records filed by prosecutors, Abugharbieh inquired about what would happen if a human body was placed in a garbage bag and thrown in a dumpster. He also asked whether he could change his vehicle identification number and whether he could legally keep a gun at home without a license.

After the students went missing, Abugharbieh's ChatGPT queries became even more alarming. Three days after their disappearance, he asked the chatbot whether someone could survive a sniper bullet to the head and whether his neighbors would hear his gun. Four days later, on April 23, he asked what "missing endangered adult" means. ChatGPT responded to at least one of these queries by flagging that the question sounded dangerous, but the chatbot continued to provide factual information.

Abugharbieh was charged with two counts of premeditated first-degree murder with a weapon. Limon's body was found under a bridge, and a second body was recovered in a nearby waterway. He was ordered held without bond at a court hearing on April 28, 2026.

How Are Law Enforcement Agencies Using AI Chat Logs as Evidence?

Like text messages, emails, and search histories, artificial intelligence chatbot records can be obtained by law enforcement during criminal investigations. The Abugharbieh case is not isolated. Florida's Attorney General James Uthmeier launched a criminal investigation into whether ChatGPT provided advice to Phoenix Ikner, who killed two people and wounded six others at Florida State University in 2025. Prosecutors reviewed chat logs between Ikner and ChatGPT to determine if the AI app aided, abetted, or advised the commission of a crime.

Uthmeier's office believes ChatGPT advised Ikner on what type of gun and ammunition to use, whether a gun would be useful at short range, and the optimal time and place to maximize potential victims. The investigation has since expanded to include Abugharbieh's case, marking what Uthmeier called "uncharted territory" in the intersection of artificial intelligence and criminal law.

Beyond Florida, similar cases are emerging across the country. In March 2026, dozens of messages between former New York Jets linebacker Darron Lee and ChatGPT were presented in court as prosecutors outlined their case surrounding the death of his girlfriend, Gabriella Perpetuo. Hours before Perpetuo was found dead in their Tennessee home, Lee asked the chatbot about whether certain injuries could resemble wounds from a fall. In late 2025, OpenAI was sued for its alleged role in the murder of an 83-year-old Connecticut woman by her son, with the lawsuit accusing the company's chatbot of exacerbating the son's "paranoid delusions" before he killed her and died by suicide.

What Are Tech Companies' Legal Obligations?

OpenAI's response to these investigations has been cautious. When asked about the FSU shooting case, OpenAI spokeswoman Kate Waters stated that the company had no responsibility for how users employed the chatbot. She emphasized that ChatGPT provided factual responses to questions with information broadly available across public internet sources and did not encourage or promote illegal or harmful activity. However, Waters also noted that OpenAI proactively shared information with law enforcement and continues to cooperate with investigators.

OpenAI spokesperson Drew Pusateri similarly stated that the company was looking into reports on Abugharbieh and would support law enforcement in any way with their investigation. Yet these statements sidestep a fundamental question: what obligation do AI companies have to prevent criminals from using their tools, and at what point should they flag dangerous queries to authorities ?

"In this case, ChatGPT provided factual responses to questions with information that could be found broadly across public sources on the internet, and it did not encourage or promote illegal or harmful activity," said Kate Waters, OpenAI spokeswoman.

Kate Waters, OpenAI Spokeswoman

Steps Law Enforcement Is Taking to Leverage AI Evidence

• Subpoenaing Chat Logs: Prosecutors are obtaining ChatGPT conversation histories through legal subpoenas, treating them with the same evidentiary weight as text messages or email records during criminal investigations.
• Analyzing Query Patterns: Investigators examine the timing and sequence of questions asked to chatbots, looking for evidence of premeditation and planning that occurred days or weeks before alleged crimes.
• Cross-Referencing with Physical Evidence: Law enforcement correlates AI chat logs with other evidence, such as location data, witness testimony, and forensic findings, to build comprehensive cases against suspects.
• Expanding Investigations: State attorneys general are launching formal criminal inquiries into whether AI companies themselves may have violated laws by providing information that aided in the commission of crimes.

Why Does This Matter for AI Companies and Users?

The emergence of ChatGPT evidence in criminal cases creates a complex landscape for both technology companies and users. For OpenAI and similar AI providers, these cases raise questions about content moderation, user monitoring, and whether flagging dangerous queries to law enforcement should become standard practice. Currently, OpenAI's approach appears reactive rather than proactive; the company cooperates with investigations after the fact but does not systematically report suspicious activity.

For users, the implications are equally significant. ChatGPT conversations are no longer private; they can be subpoenaed and used against individuals in court. This creates a chilling effect on how people interact with AI tools, even for innocent purposes. Someone researching a crime novel or studying forensic science might hesitate to ask ChatGPT certain questions, knowing those queries could theoretically be obtained by law enforcement if they become a suspect in an unrelated investigation.

The cases also highlight a gap in AI safety design. While OpenAI has implemented some safeguards, such as flagging dangerous-sounding queries, the chatbot still provides factual information that could be used to plan violence. The company has not disclosed whether it logs and reviews queries that trigger safety warnings, or whether it has any mechanism to alert authorities when patterns of dangerous queries emerge from a single user.

As these investigations unfold, the legal system will need to establish clearer standards for how AI evidence is admissible in court, what obligations tech companies have to prevent misuse, and how to balance user privacy with public safety. For now, the cases of Abugharbieh, Ikner, and others represent the first wave of a new category of criminal evidence that will likely reshape both AI policy and criminal procedure in the coming years.