OpenAI faces probe and tweaks

Florida’s attorney general just opened a state investigation into OpenAI, and he did it while promising subpoenas before he has spelled out the full scope of the case. Politico reported on April 9 that Attorney General James Uthmeier tied the probe to public-safety and national-security concerns around ChatGPT. (politico.com) The immediate spark was a new wave of allegations around the 2025 Florida State University shooting. Politico said lawyers for a victim’s family claim the gunman was in “constant communication” with ChatGPT and that the chatbot may have helped him plan the attack, while OpenAI said it would cooperate. (politico.com) This is not the same kind of pressure OpenAI has been dealing with in court. The New York Times copyright case survived a motion to dismiss in March 2025, and the Federal Trade Commission published a report in January 2025 warning that cloud-and-artificial-intelligence tie-ups like Microsoft and OpenAI can create lock-in and competition risks. (npr.org) (ftc.gov) Florida’s move is different because it comes from a state political office and focuses on harms from the product itself, not just who owns what data or who controls the market. Politico said Uthmeier also asked the Legislature to expand his authority over artificial intelligence after statewide regulation efforts under Governor Ron DeSantis stalled. (politico.com) At the exact same time, OpenAI has been changing the product in quieter ways that most users only notice when they hit a limit. OpenAI’s model release notes say that since March 18, paid ChatGPT users who run out of GPT-5.4 Thinking capacity are automatically shifted to GPT-5.4 mini, a smaller fallback model that keeps the chat going instead of cutting it off. (help.openai.com) That fallback model is basically the spare tire in the trunk. OpenAI says GPT-5.4 mini is not shown in the model picker for paid users, but it takes over after rate limits so reasoning chats can continue during heavy usage. (help.openai.com) OpenAI also changed the price ladder on April 9. In its developer community announcement, the company said it added a new $100-a-month Pro tier with 5 times more Codex usage than Plus, while keeping a $200 Pro option for heavier users with 20 times the Plus limits. (community.openai.com) That pricing change tells you where OpenAI thinks demand is moving. The company said the new tier is meant for “longer, high-effort Codex sessions,” which means people are using ChatGPT less like a search box and more like a contract worker that writes, debugs, and revises code for hours at a time. (community.openai.com) So the same week brought two very different signals about the future of artificial intelligence. One came from Tallahassee in the form of subpoenas and safety claims, and the other came from OpenAI in the form of hidden reliability tweaks and a cheaper middle rung for power users. (politico.com) (help.openai.com) (community.openai.com)