OpenAI Managing Director and former Chancellor of the Exchequer George Osborne said governments and the public need to scrutinise how frontier AI behaves, including when it goes wrong, and that companies building these systems have a responsibility to share evidence. He spoke after Anthropic top safety researcher Evan Hubinger warned in a post on X that, as AI advances quickly, there is a greater than 10% chance it "could kill all humans" within the next decade, saying risk from currently existing models was "low" but he was "worried" the technology might soon improve itself to pose an existential risk. In a LinkedIn post, Osborne said OpenAI already publishes findings about model misalignment, and its new reporting framework sets out a more systematic approach to tracking, investigating and disclosing it, aiming to publish findings sooner even when they are not fully explained or resolved. He said alignment and monitoring remain difficult, evolving challenges, and sharing findings lets researchers, policymakers and others examine evidence, test assumptions and improve safeguards, with the framework intended to help establish clearer industry disclosure standards developed with other AI companies, researchers, standards bodies and regulators, refined through experience and public feedback.