OpenAI, Anthropic, Meta and Google representatives stopped short of guaranteeing their AI agents will always obey safeguards during a New York City Council hearing.