Anthropic and OpenAI push for AI safety rules as they shape regulation

Anthropic and OpenAI CEOs are publicly warning that advanced AI systems pose dangers requiring regulation and independent testing, a stance experts say may serve their political and financial interests ahead of U.S. midterm elections and potential stock market listings. The companies are shaping safety discourse to position themselves as cautious leaders in an unregulated market. The firms advocate for self-chosen auditors and voluntary protocols rather than oversight from existing federal agencies, while critics note this focus on existential risks diverts attention from current harms like AI-enabled surveillance and military use. Former government evaluators question the independence of embedded evaluators, and analysts see the rhetoric as creating barriers against smaller competitors. President Donald Trump has dismissed AI risk warnings as a hoax, and his administration has not pursued new regulations. Meanwhile, some former employees like Daniel Kokotajlo argue the companies' calls for caution dissipate political will for meaningful action, as AI systems continue advancing rapidly with unresolved safety concerns.