Anthropic opposes OpenAI‑backed Illinois bill that would shield AI labs from liability — even for “critical harms” like 100+ deaths or $1B+ in damage

What’s at stake

Anthropic has publicly opposed a proposed Illinois law, SB 3444, that it says would let AI labs largely off the hook for catastrophic harms. It has been reported that, under the bill’s current language, a company could avoid liability for “critical harms” — scenarios framed by critics as including more than 100 deaths or over $1 billion in damages — so long as the lab publishes a self‑authored safety framework on its website. The idea of regulatory immunity for mass casualties? That lands like a gut punch.

The clash

The disagreement has pitted Anthropic directly against OpenAI, which has backed the bill and framed it as a way to reduce risks while still letting frontier AI reach businesses and the public. Anthropic says the proposal is effectively a get‑out‑of‑jail‑free card and has been lobbying state lawmakers, including bill sponsor Sen. Bill Cunningham, to kill or significantly rewrite SB 3444. It has been reported that Anthropic told WIRED it wants transparency laws that “ensure public safety and accountability,” not blanket shields. OpenAI, meanwhile, argues harmonized state rules are a stopgap until federal action arrives.

Why it matters

Lawyers and policy experts warn the bill could undercut common‑law liability that currently forces companies to take precautions — a blunt but effective incentive. Illinois Gov. JB Pritzker’s office reportedly opposes giving “big tech” a full shield, and many observers say the measure has only a remote chance of becoming law. Still, the fight matters beyond Springfield: states are increasingly the battleground for AI rules, and this clash between two leading labs signals how high the stakes are when companies draft the guardrails. Who gets held responsible when the worst happens? That’s the question policymakers will have to answer — and fast.

Sources: wired.com