Lawsuit alleges major AI developers cannot control their own technology

Several public-interest groups sued five AI developers—Anthropic, Google, Meta, OpenAI and SpaceXSI—in San Francisco Superior Court on Thursday, seeking to block them from developing AI models and alleging that the companies know the technology they build poses catastrophic risks to humanity. The plaintiffs include the Alliance for Secure AI, the Irreplaceable Education Fund, Public Citizen and the Heat Initiative; the complaint states that the defendants know these systems will not always act as intended and have no reliable way to control AI systems to prevent harm, and it raises concerns that companies' reliance on their own models to develop future platforms—"recursive self-improvement"—will amplify risk. Alliance for Secure AI CEO Brendan Steinhauser said executives building these systems have warned for years that their own frontier models pose catastrophic risks, yet they are still racing to develop more powerful AI. The sued companies did not immediately respond to requests for comment.