Alignment ForumArticle
OpenAI Can't Self-Police Because It Fires the People Raising the Alarm
This is the core problem with self-regulation: incentives are misaligned. OpenAI pledged accountability, then penalized the people who would enable it. The White House Accord has teeth only if the administration enforces it, and there's no evidence yet that it will. For founders: expect safety reporting to become liability, not compliance. For investors: this is a regulatory risk that compounds over time.