OpenAI, Anthropic, Google have been in talks on AI safety for weeks | Dark Hacker News