Legal heat on OpenAI

Florida is trying to pull OpenAI into a criminal-style probe at the same moment OpenAI is backing a bill in Illinois that would make it harder to sue powerful artificial intelligence labs over the worst kinds of damage. Attorney General James Uthmeier said on April 9 that Florida is investigating whether ChatGPT and other OpenAI products have contributed to harm involving minors or criminal activity. (fox35orlando.com) The Florida move did not come out of nowhere. It followed a planned lawsuit tied to the April 2025 Florida State University shooting, where authorities allege the gunman used ChatGPT to research details such as the busiest times at the student union. (fox35orlando.com) Uthmeier said subpoenas are coming, which means this is not just a speech or a warning letter. His office is looking at records and internal practices, even though the exact scope of the investigation has not been spelled out yet. (fox35orlando.com) Florida is framing the case around children as well as public safety. Uthmeier said his office is examining whether OpenAI’s systems may harm minors, and he called on the Florida Legislature to add protections for children and give the attorney general more power to act on artificial intelligence harms. (fox35orlando.com) Two days before that probe became public, OpenAI published a Child Safety Blueprint on April 8. The company said the document was built with feedback from the National Center for Missing and Exploited Children, Thorn, and the Attorney General Alliance to fight artificial intelligence-enabled child sexual exploitation. (openai.com) Then came the Illinois fight. OpenAI publicly backed Illinois Senate Bill 3444, called the Artificial Intelligence Safety Act, which would protect developers of “frontier” models from liability for “critical harms” if the company did not intentionally or recklessly cause the harm and if it published safety, security, and transparency reports on its website. (letsdatascience.com, legiscan.com) The bill uses a very high bar for the systems it covers. A “frontier model” is defined as one trained with more than $100 million in computing costs, which points at the biggest labs rather than small software firms. (letsdatascience.com) The harms in that bill are also very large. The thresholds include 100 or more deaths or serious injuries, at least $1 billion in property damage, or helping a bad actor develop a chemical, biological, radiological, or nuclear weapon. (letsdatascience.com) OpenAI’s argument is that rules like this push companies to publish safety paperwork without freezing the technology. Spokesperson Jamie Radice said the company supports approaches that target serious harm from the most advanced systems while still letting businesses and consumers use the tools. (letsdatascience.com) Critics see the same bill as a shield built before the first giant courtroom test arrives. Illinois bill text says the law would stop applying if the federal government later passes overlapping rules, which makes the state fight look like a race to set the template before Washington does. (ilga.gov, letsdatascience.com) That is why these two stories fit together. In Florida, prosecutors are asking whether OpenAI did enough to keep young users and the public safe; in Illinois, OpenAI is asking lawmakers to narrow when the company can be blamed if a powerful model is tied to catastrophe. (fox35orlando.com, letsdatascience.com)