Anthropic CEO Dario Amodei called the exponential pace of AI progress a warning sign. He said the industry had long lied about the technology’s risks. He now urges slowing AI development. He supports federal regulation and a potential AI kill switch. He opposes an outright ban on superintelligence. He cited nuclear arms agreements as a possible model for international coordination. He expressed discomfort that private companies control such powerful technology. He proposed joint governance by democratically elected governments. He said proper testing of each model generation is essential. He will give independent evaluators permanent access to Anthropic’s models. He warned that building AI the wrong way greatly increases the risk of harm.