AI labs split on liability
Investors are raising questions about OpenAI’s $852bn valuation as the company shifts focus toward enterprise customers. (reuters.com) At the same time Anthropic publicly opposed an Illinois AI liability bill that OpenAI backed, highlighting a growing divide among leading labs over legal responsibility for harms. (wired.com) Those developments are feeding a wider industry debate that the next phase may hinge less on raw model progress and more on who pays and who is held liable when systems cause serious harm. (reuters.com)
OpenAI’s business model and the rules for who pays when artificial intelligence systems cause harm are now colliding in public. (reuters.com) The Financial Times reported on April 14 that some investors are questioning OpenAI’s $852 billion valuation as the company leans harder into enterprise customers, a market with slower sales cycles and heavier demands around security, reliability, and legal risk. Reuters cited that report and said the debate centers on whether revenue growth can match the price investors are assigning to the company. (reuters.com) At the same time, Wired reported on April 14 that Anthropic opposed an Illinois bill that OpenAI backed, splitting two leading labs over liability for severe harms tied to advanced models. The bill, Senate Bill 3444, was referred to assignments in the Illinois Senate on February 4, 2026, according to the Illinois General Assembly. (wired.com; ilga.gov) Liability is the legal question of who can be sued when a product causes damage. In artificial intelligence, that can mean the model developer, the company that builds an app on top of the model, or the customer that uses the system in the real world. (wired.com; reuters.com) Illinois is not setting national law, but states have become testing grounds while Congress has not passed a broad artificial intelligence liability framework. That leaves labs, app makers, and corporate buyers watching state bills for signals about how courts and lawmakers may divide responsibility. (wired.com; ilga.gov) OpenAI’s push toward enterprise customers puts that question closer to its core business. Large companies buying artificial intelligence tools usually ask for contracts that spell out indemnity, warranties, security terms, and who absorbs losses if a system fails or is misused. (reuters.com) Anthropic’s break with OpenAI also cuts against the recent habit of major labs presenting a shared front on safety policy. The split suggests that once legislation moves from broad safety principles to specific legal exposure, labs’ commercial incentives can diverge. (wired.com) Illinois has several artificial intelligence bills in play this session, including House Bill 3506, the Artificial Intelligence Safety and Security Protocol Act, which would require developers to publish and follow safety and security protocols. That wider state-level activity gives the liability fight more weight than a single stand-alone dispute. (ilga.gov; ilga.gov) For investors, lawmakers, and customers, the argument is no longer only about which lab has the strongest model. It is also about which company can sell those systems at scale while defending the legal responsibilities that come with them. (reuters.com; wired.com)
