Anthropic opposes AI liability bill backed by OpenAI

Anthropic is opposing Illinois bill SB 3444. The bill, backed by OpenAI, would shield AI labs from liability for large-scale harm caused by misuse of their models, as long as they draft and publish their own safety frameworks.

People familiar with the matter said Anthropic has urged state senator Bill Cunningham and other lawmakers to change or drop the bill.

Governor JB Pritzker’s office said he does not support giving big tech a full shield from responsibility, while OpenAI said the measure would reduce risk and support a harmonized approach to state AI rules.

Critics, including Secure AI Project’s Thomas Woodside, said the bill could nearly eliminate existing common law liability.

Anthropic last week supported a separate Illinois bill that would require public safety plans and third-party audits for frontier AI developers.

🔗 Source: Wired

🧠 Food for thought

Implications, context, and why it matters.

The battle over AI liability is part of a bigger national script

Anthropic’s stance in Illinois matches its work in other states, where it has backed laws that push transparency and accountability 1.
Anthropic was the first major AI lab to publicly support California’s SB 53, which requires covered companies to disclose safety and security protocols and report critical safety incidents to state officials 2.
The “trust but verify” model later appeared in New York’s RAISE Act, a state AI law, which helped set a bicoastal standard 3.
OpenAI has supported Illinois’ SB 3444, which follows a different regulatory path and sits alongside well-funded AI lobbying in other states 4, 5.

AI labs are creating the patchwork of state laws they claim to oppose

OpenAI has argued for unified federal rules to avoid a “patchwork of inconsistent state requirements,” yet it has still pushed state policy, including support for Illinois’ SB 3444 4.
Large AI labs are also promoting competing playbooks, with Anthropic favoring disclosure rules in California while OpenAI supports liability limits in Illinois 1, 4.
The split leaves states with different legal rules. New York’s RAISE Act calls for reporting critical safety incidents within 72 hours, while Illinois’ proposed SB 3444 would protect AI developers from liability for certain “critical harms” as defined in the bill when listed conditions apply 4, 6.
These state fights operate as a policy testing ground. Their results may shape future federal AI legislation through precedent.