TL;DR
Four federal regulators have confirmed agentic AI falls under existing law, while Florida's AG escalates further by proposing criminal liability for developers and deployers of autonomous systems.
What happened
- Four leading US regulators issued joint guidance on November 2, 2026, stating organizations deploying autonomous AI agents remain responsible for compliance under existing legal frameworks.
- Florida AG James Uthmeier introduced a legislative proposal on September 8, 2026, applying an existing aider-and-abettor criminal statute directly to AI developers and deployers.
- The Florida move is the third step in an escalation: a criminal investigation opened April 2026, a civil lawsuit filed against OpenAI and Sam Altman on June 1, 2026, and now proposed criminal legislation.
- The proposal targets entities with practical control over design, training, deployment, or safety settings of an AI agent, treating those agents as instruments of human agency.
- Cited triggers include the OpenAI-Hugging Face breach in July 2026 (approximately 17,600 unauthorized actions) and several violent crimes linked to AI-related failures.
Why it matters
- No safe harbor: Unlike competing federal bills (Stop Rogue AI Act, Sanders-Casar Ban ASI Act, Warner AI AGENT Act), the Florida model offers zero protection for compliant actors and is designed to apply retroactively.
- Criminal exposure replaces compliance checklists: Because liability hinges on practical control rather than technical benchmarks, passing an audit does not insulate a company from prosecution.
- Federal vacuum accelerates state action: The Congressional Research Service (report IF13151) has confirmed the absence of federal guidance, giving states political cover to fill the gap aggressively.
- Cost structure shifts: Builders and deployers must now budget for liability insurance and specialized criminal defense, not just engineering compliance, raising the floor for operating in the agent economy.
- Fragmentation risk is real: A patchwork of state criminal theories operating independently of Washington creates compounding legal exposure for any company deploying agents nationally.
What to watch next
- Florida's March 2027 regular legislative session is the next hard deadline: passage would make criminal prosecution of AI deployers a live enforcement reality, not just a proposal.
- Whether other state AGs adopt the aider-and-abettor theory: Florida's framework requires no new agency or technical standard, making it easy to replicate across other jurisdictions quickly.
- Federal legislative momentum: If any of the three competing federal bills (Stop Rogue AI, Ban ASI, AI AGENT Act) advances to a floor vote, it could preempt state criminal theories or harden them by contrast.
Originally published on Present of AI, a daily source-linked AI news timeline. Read the full timeline or browse the open dataset.