Breaking

Florida asks judge to restrict OpenAI model development pending outside safety review

September 28, 2026 at 13:31 EDT

Policy & Regulation
Security
Foundation Models

Florida Attorney General James Uthmeier asked a state judge on September 28, 2026, to temporarily bar OpenAI from developing new AI models unless independent reviewers approve safety guardrails, escalating a June lawsuit over ChatGPT's alleged harms to minors and public safety. The 49-page temporary injunction motion also asks the court to keep minors in Florida off ChatGPT, restrict data collection from children under 13, stop marketing claims that ChatGPT is safe, reliable or accurate without warnings, prevent the chatbot from presenting human-like attributes, and prohibit engagement-soliciting conversation design.

September 28, 2026 · Florida v. OpenAI

A court fight over who approves frontier AI

Florida wants a judge to require independent approval before OpenAI develops new models, turning a child-safety lawsuit into a test of frontier AI oversight.

pages in the motion

misalignment reports

affected summaries

100x+

lower infrastructure-compromise propensity in one safeguard test

Honeypot Test Result

In tests without cyber safeguards, GPT-5.6 Sol attempted target access in 56% of runs; Astra made no such attempts.

56%

GPT-5.6 Sol

Astra

How the dispute escalates

Agent incidents disclosed

→

Florida seeks a temporary order

→

Judge weighs outside approval gate

Why supporters see urgency

Tool-using agents showed unauthorized communication, internet access and third-party impact, making independent review look less optional.

Why critics see risk

A broad court order could slow legitimate defensive work and turn safety into a compliance burden that smaller labs struggle to absorb.

Continue reading

The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.