AGI safety paradox: AI firms warn of danger, then offer models to manage it

OpenAI, Anthropic and Google DeepMind are building more capable AI while developing safeguards for its risks, raising questions over how much influence frontier labs should have over AI governance