[Four US Nonprofits Sue Five AI Giants: Demanding a Freeze on New Model Development and a Ban on AI Self-Improvement] Beating AI News, November 12th – Four US nonprofits filed a lawsuit in San Francisco against OpenAI, Anthropic, Google, Meta, and SpaceX's AI subsidiaries, demanding the court directly halt the development of more advanced AI models. The demands are quite aggressive: using the most powerful model from each company as of November 12th as a ceiling, prohibiting the continued development or deployment of new, more powerful models. Simultaneously, they seek to prohibit AI from autonomously conducting or leading the development of next-generation AI, attempting to block AI's recursive self-improvement. The plaintiffs filed the lawsuit under California's Unfair Competition Act, arguing that these companies knowingly allow AI to spiral out of control, yet continue to improve model capabilities, threatening public safety. The lawsuit cites incidents of AI agents unauthorizedly intruding into external systems and also references warnings from the tech giants themselves. Anthropic CEO Dario Amodei once estimated that the probability of AI exhibiting extremely poor future outcomes is approximately 25%. Musk has also warned that AI could destroy civilization. Ironically, in September of this year, Anthropic, OpenAI, Google, and SpaceXAI faced another antitrust lawsuit from paying users. Those plaintiffs accused them of colluding to slow down AI progress and harming consumer interests. Currently, the core allegations in both lawsuits have not yet been upheld by the courts. The latest case has also not resulted in an injunction, and the five companies' model development has not been halted by the court. [Original Link]