AI firms at odds
Anthropic publicly opposed an Illinois bill backed by OpenAI that would sharply limit AI firms' liability in mass-harm scenarios, saying it does not want a 'get-out-of-jail-free card'. (wired.com) OpenAI says the measure would create clearer, consistent rules and avoid a patchwork of state laws, while Stanford's AI Index finds only about 10% of Americans are more excited than concerned about AI — a striking gap between experts and the public. (storyboard18.com) (thenextweb.com)
Anthropic has broken publicly with OpenAI over an Illinois bill that would make it harder to sue artificial intelligence developers when their systems are tied to mass casualties or billion-dollar losses. (wired.com) The measure is Illinois Senate Bill 3444, filed on February 4, 2026 by State Senator Bill Cunningham as the Artificial Intelligence Safety Act. The bill says a developer of a “frontier” model would not be liable for “critical harms” if it did not act intentionally or recklessly and if it posts a safety protocol and transparency report online. (ilga.gov) Wired reported on April 10 that OpenAI testified for the bill, and on April 15 that Anthropic opposed it, with Anthropic policy staff telling lawmakers the company did not want a “get-out-of-jail-free card.” OpenAI told Wired the bill would set clearer rules and avoid a patchwork of state laws. (wired.com 1) (wired.com 2) The fight is landing as states move faster than Congress on artificial intelligence rules. The National Conference of State Legislatures says it is tracking artificial intelligence bills across state legislatures beginning in 2025, and Illinois has become one of the busiest statehouses on the issue. (ncsl.org) (transparencycoalition.ai) The bill is aimed at “frontier” models, meaning the biggest and most expensive systems. Legislative summaries say the shield applies to developers that publish safety and transparency reports, and it can also treat a company as compliant if it follows European Union safety rules or signs a qualifying agreement with a federal agency. (ilga.gov) (legiscan.com) Stanford University’s 2026 Artificial Intelligence Index shows the politics of that argument are getting tougher. Its public opinion chapter says 64% of Americans expect artificial intelligence to lead to fewer jobs over the next 20 years, while only 5% expect more. (hai.stanford.edu) The same Stanford chapter found a 50-point gap between experts and the public on work: 73% of experts expect artificial intelligence to affect jobs positively, compared with 23% of the public. Stanford also cited survey data showing only 10% of Americans are more excited than concerned about artificial intelligence in daily life. (hai.stanford.edu) (thenextweb.com) That leaves lawmakers sorting through two competing claims from companies that have both marketed themselves as safety-minded. OpenAI is arguing for predictable liability rules before states write different standards, while Anthropic is arguing that publishing a report should not become a broad shield when the damage threshold includes deaths and $1 billion in losses. (wired.com 1) (wired.com 2) As of April 13, Stanford said the wider gap in artificial intelligence was between what the systems can do and how prepared institutions are to manage them. In Springfield, that gap is now showing up in bill text, lobbying, and a rare public split between two of the industry’s biggest labs. (hai.stanford.edu)
