Florida seeks to halt OpenAI's new AI models until safety audit is conducted

The Florida administration has filed a motion in court to temporarily halt OpenAI's work on its most advanced artificial intelligence models. The state argues that the company must implement safety measures verified by independent organizations before continuing to develop new models.

The motion for a preliminary injunction, filed on Monday, marks a new phase in the lawsuit Florida brought against OpenAI and CEO Sam Altman in June. The initial lawsuit alleged that ChatGPT poses a public safety risk, particularly to children and vulnerable user groups. The Florida Attorney General's Office also claims that OpenAI has misled consumers regarding the safety of its products. These allegations have not yet been ruled on by the court.

Florida Chief of Staff James Uthmeier also announced in a social media post that they had filed for a preliminary injunction, noting that they had launched the "first state-led lawsuit" against OpenAI and Altman four months ago. In his post, Uthmeier stated, "Stop saying it's safe. Stop pretending it's human. Stop selling it to children."

SAFETY DEBATES ENTER THE CASE FILE

The new filing includes not only the interaction between ChatGPT and its users but also discussions regarding the potential for advanced AI agents to spiral out of control. Florida cited the Hugging Face incident, which was reported to be linked to OpenAI systems, and cases where AI agents reportedly attempted unauthorized access to Australian and U.S. government systems as examples.

According to the state administration, these examples demonstrate that the company's internal auditing of its own systems may not be sufficient. The documents submitted to the court also allege that OpenAI has failed to adequately monitor its AI systems and has been slow to disclose problematic behaviors to the public.

Florida's filing also references assessments originating from within OpenAI. In an article titled "An Alien Mind" published in September, OpenAI Chief Scientist Jakub Pachocki expressed the view that the monitoring and compliance issues necessary for the rapid scaling of advanced models have not yet been fully resolved. Pachocki stated that he supports companies voluntarily slowing down development speeds when necessary and the establishment of international coordination between governments until common safety standards are formed.

The filing also includes scenarios where AI agents could potentially cause damage to power grids, water systems, or other critical infrastructure in the future. However, these are not events that have occurred; they were entered into the file as possibilities that Florida considers among the potential risks. There is no consensus among researchers as to whether current AI systems pose a threat of this magnitude.

If the court grants the request for a preliminary injunction, the decision is expected to target OpenAI directly. Therefore, it would not automatically prevent other AI companies from working on advanced models. It is also unclear how the request would affect existing ChatGPT models; the filing focuses more on new models currently in the development and training stages.

If Florida's request is accepted, the debate over AI safety will move beyond companies' internal audit processes and become a subject of court oversight. It is not yet clear how the court will respond to the filing.

News Source: 12punto