OpenAI faces probe in Zamil-Bristy murder case

The US state of California’s Florida has expanded its criminal investigation into artificial intelligence firm OpenAI to include the killings of Zamil Limon and Nahida Bristy, after prosecutors revealed the suspect allegedly used ChatGPT in connection with the crime.

The announcement came days after Tampa prosecutors disclosed that the primary suspect, Hisham Abugharbieh, 26, had allegedly consulted ChatGPT about disposing of human remains.

Abugharbieh is charged with two counts of first-degree murder with a weapon, along with additional offences including false imprisonment and unlawfully handling a dead body.

“We are expanding our criminal investigation into OpenAI to include the USF murders after learning the primary suspect used ChatGPT,” Florida Attorney General James Uthmeier said in a statement posted on social media.

The case centres on the deaths of 27-year-old Bangladeshi doctoral students Zamil Limon and Nahida Bristy, who were reported missing on April 16.

Limon’s body was found on April 24 in a trash bag near the Howard Frankland Bridge, while human remains believed to be Bristy were recovered from a Tampa waterway over the weekend. Authorities have not yet formally confirmed the second victim’s identity.

New court filings submitted to oppose bail for Abugharbieh include excerpts of alleged conversations between the suspect and the AI chatbot.

Investigators say the exchanges, spanning several days before the students’ disappearance to hours before Limon’s body was discovered, included questions about placing a human body in a garbage bag, whether a person could survive a gunshot wound to the head, and whether a vehicle identification number could be altered.

Other queries reportedly included whether vehicles are checked at a state park during early morning hours, whether neighbours could hear a gunshot, and the meaning of a “missing endangered adult” designation.

Authorities allege that Abugharbieh, who was Limon’s roommate, was arrested hours after the first body was discovered following a confrontation with deputies.

Family members of the victims have expressed grief and called for justice.

Zubaer Ahmed, Limon’s brother, described the loss as devastating.

“We are in a devastated situation and Jamil is not just any student,” he said. “He is my brother, a child of my mother.”

“We just want justice and accountability, because it’s not only about Zamil or Nahida -- it is about all international students,” he added.

The case has intensified scrutiny of generative AI tools and their safeguards, particularly their potential misuse in criminal activity.

It follows a separate investigation launched by Uthmeier into OpenAI’s role in a deadly shooting at Florida State University last year.

In that case, prosecutors said the accused gunman had used ChatGPT shortly before the attack to ask about firearm use, campus activity patterns, and how many casualties would attract national media attention.

Uthmeier’s probe is considered the first attempt in the United States to examine whether AI companies or developers could face criminal liability for how their technology is used.

His office is now weighing whether to issue a formal subpoena to OpenAI in the USF case, similar to the action taken in the Florida State investigation.

Experts say assessing the role of AI in such incidents is complex.

Dr Jill Schiefelbein, an AI expert and adjunct professor at the University of South Florida, noted that systems like ChatGPT retain limited interaction histories to detect misuse.

“It has that 30-day memory, so safeguards can flag potentially illegal uses and send them to the appropriate channels,” she said.

However, she cautioned that determining whether such safeguards are sufficient remains an evolving challenge.

“Unfortunately, that can only be known through trial and error. That’s why technology companies actively test these guardrails,” she added.