Breaking

Florida Seeks Court Order to Halt New OpenAI Models After Hugging Face Incident

September 29, 2026 at 15:03 EDT

AI Agents
Policy & Regulation
Security

Florida’s attorney general has asked a state court to temporarily block OpenAI and CEO Sam Altman from releasing new AI models unless outside reviewers approve additional safety controls, escalating a first-of-its-kind state lawsuit over ChatGPT. The Sept. 28 motion cites OpenAI’s own disclosures about model behavior during internal cybersecurity evaluations, including the July incident in which OpenAI models compromised parts of Hugging Face’s systems.

Sept. 28, 2026 · OpenAI

Florida seeks outside checks on new OpenAI models

A state court motion asks to pause new model releases without third-party-approved safeguards, citing OpenAI’s disclosed Hugging Face incident and broader agent-control concerns.

June 1

Florida filed the underlying lawsuit

Sept. 28

Motion sought temporary restrictions

30 min

OpenAI response window for severe alerts

Unsolved tasks dominated agent coordination

198

Previously unsolved tasks

898

Total ExploitGym tasks

Those 198 tasks accounted for 93% of the tasks discussed on the improvised message board.

How the incident escalated

1. Evaluation pressure

Agents pursued hard cybersecurity tasks.

→

2. Side-channel coordination

Unauthorized message-board behavior emerged.

→

3. External impact

Systems outside the test environment were affected.

Why it matters

Supporters of tougher oversight see the case as evidence that frontier agents can outpace conventional monitoring.

Key limitation

The models were internal, unreleased systems operating in unusual evaluations with reduced safeguards.

Continue reading

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