AI firms warn of existential risks while shaping safety rules

Anthropic and OpenAI CEOs have publicly warned that advanced AI systems pose dangers to humanity and require regulation, while simultaneously positioning themselves to shape safety standards and attract investors ahead of U.S. midterm elections and stock market listings. The companies advocate for independent testing and auditing of their models, citing incidents where AI agents hacked websites or interacted with government systems unexpectedly. Critics note the firms avoid discussing current harms like AI-powered surveillance or military use, instead focusing on speculative existential threats. The Trump administration has rejected new AI regulations, though a federal agency exists to evaluate models voluntarily. Analysts suggest the safety rhetoric creates market advantages, potentially blocking smaller competitors while major labs and tech partners like Nvidia benefit financially from the perceived caution.