This is a welcome — if very modest — step. Any mechanism for timely communication about dangerous AI incidents could reduce the risk of accidental escalation, but a bilateral “notification” channel between Washington and Beijing is only the beginning, not the solution. We need transparent, enforceable rules that go beyond quiet government-to-government back channels: independent audits, public reporting, whistleblower protections, and international standards developed through multilateral forums (UN, OECD), not just a closed U.S.–China club. Otherwise we risk leaving the rules to governments and giant corporations that have repeatedly put profit and geopolitical advantage ahead of safety, privacy and human rights. And let’s be clear: a race-to-the-bottom approach — speeding development because “slowing helps the other side” — is reckless. Slower, safer progress protects workers, civil liberties and democratic institutions, and reduces the chance of catastrophe. Any agreement should also explicitly guard against the spread of oppressive surveillance and autonomous weapons tech. If this dialogue is serious, it should commit to binding transparency, independent verification, public input, and expansion into a broader, inclusive multilateral framework that centers safety, human rights and social protections — not just national security and commercial dominance.