OpenAI is backing Illinois bill SB 3444 that would shield AI labs from liability when their models are used to cause mass casualties (100+ deaths) or financial disasters ($1+ billion in damage), as long as companies publish basic safety reports. Anthropic opposes the legislation, with head of government relations Cesar Fernandez calling it a "get-out-of-jail-free card" that prioritizes transparency over real accountability. The disagreement centers on whether AI developers should bear responsibility when bad actors weaponize their models — like using AI to create bioweapons that kill hundreds.
This fight reveals a fundamental split in how leading AI companies want to approach regulation as they scale up lobbying efforts nationwide. OpenAI has shifted from playing defense against liability bills to actively promoting ones that limit their exposure, part of what they call a "harmonized" regulatory approach across states. The $100 million training cost threshold in the bill would capture all frontier labs — OpenAI, Anthropic, Google, Meta, and xAI — but only if they can prove they didn't "intentionally or recklessly" cause harm.
The timing isn't coincidental. This follows my previous reporting on Anthropic's Pentagon disputes and their push for stricter AI governance frameworks. While policy experts say SB 3444 has slim chances of passage, it's setting precedent for how AI companies want liability structured. Anthropic is lobbying Senator Bill Cunningham directly to either kill the bill or substantially revise it, suggesting they see this as a critical battleground for future AI regulation.
For developers and AI users, this debate will shape what safeguards get built into production systems versus what gets handled through legal frameworks. If OpenAI's approach wins, expect more emphasis on compliance documentation and less on technical safety measures that actually prevent misuse.