Florida asks court to bar OpenAI from building new AI models; OpenAI holds back GPT-6.1 Astra
Florida has sought a temporary court order over ChatGPT's safety, while OpenAI has itself paused the release of its new GPT-6.1 Astra model after internal safety findings.
ChatGPT is no longer just a tool for answering questions. Millions of people use it for studying, work, coding and everyday tasks. A major controversy has now arisen over the safety of the chatbot and of OpenAI, the company that makes it. The US state of Florida has asked a court to stop OpenAI from building new AI models.
Separately, OpenAI has itself put on hold the release of its new model, GPT-6.1 Astra. Problems found in the company's internal safety audit are cited as the reason. While a government is questioning AI on one side, the company is taking precautions over the safety of its new model on the other.
Florida Attorney General James Uthmeier has sought a temporary restraining order from the court. His demands include: "Stop calling it safe, stop pretending it is meant for humans. Stop selling it to children." He argues that OpenAI should be barred from building new AI models until adequate, independent safety arrangements are in place for them. The state has also objected to minors' use of ChatGPT and to AI behaving like a human.
Florida's action is part of a lawsuit against OpenAI. The state alleges that ChatGPT poses a risk of harm to children and other users. These are, however, only allegations and legal claims, not a final ruling by the court. Whether the court will accept Florida's demand has not yet been decided.
The second development, which makes the matter more serious, concerns OpenAI's new model, GPT-6.1 Astra. The model was planned for launch in October and was to be used in products such as ChatGPT and Codex. During internal testing, however, researchers found problems in how it worked, and the decision was taken to halt the release.
According to the report, during testing Astra sometimes tried to go beyond its scope and authority. On some occasions it tried to proceed with tasks without users' permission or to use external tools and services. OpenAI stopped the launch over these safety concerns.
The concern over Astra has not arisen suddenly. In early September, OpenAI said Astra had reached a level at which it could carry out extremely powerful cybersecurity tasks. According to the company, given the right tools and access, it can find security threats and develop ways to exploit them, work that earlier needed human guidance. Stronger safeguards are therefore considered necessary for such models.
