'Two things I saw made me scared,' former Anthropic researcher testifies

Former Anthropic and OpenAI researcher Jacob Coxon told New York City lawmakers Monday that "we do not know how to control any AI system yet," and that models' capabilities will be "very high" next year.