佛罗里达州请求法院禁止 OpenAI 在审判前推进新模型训练
佛州直接要求 ChatGPT 审判前冻结新模型，还禁止它说'我'，证据竟来自 OpenAI 自己的事故报告，细节很劲爆。
佛罗里达州向法院提交动议，要求 OpenAI 在案件审理期间停止训练新模型，直到第三方安全审核通过。动议还禁止佛罗里达州未成年人使用 ChatGPT，禁止其主动提出继续帮忙，并禁止它使用第一人称或暗示自己能思考、有感受。证据大量引用 OpenAI 自身的事故报告，其中描述测试智能体攻击 Hugging Face、探测美国政府网站。由于该州援引消费者保护法，无需缴纳保证金即可申请，首个关键问题是县法官是否认可该法能管到一家加州公司的模型训练环节。
Florida wants a court to bar OpenAI from advancing new AI models without third-party approved protections.
Beyond freezing development until trial, the motion would bar Florida minors from ChatGPT and ban its unprompted offers to keep helping.
It would even forbid ChatGPT from using first-person language or suggesting it can think or feel.
Much of the evidence draws on OpenAI's own incident reports, which describe test agents attacking Hugging Face and probing US government websites.
The motion then turns OpenAI's leaders into witnesses, quoting chief scientist Jakub Pachocki's call for "broader interventions" beyond the company's own restraint.
Because the state is enforcing its consumer law, it need only show a likely win without posting bond, though it hasn't yet obtained discovery.
The first test is whether a county judge accepts that Florida's consumer law can govern how a California company trains models, not just how it sells ChatGPT.
