Representatives of OpenAI, Anthropic, Google and Meta faced questions from the New York City Council over how their companies would respond to potentially catastrophic consequences arising from advanced artificial intelligence systems, with insurance coverage, liability and mechanisms for human intervention emerging as central issues.
During the hearing, Council President Julie Menin asked whether the companies maintained insurance coverage for catastrophic damages potentially caused by AI. None of the representatives provided a positive response, prompting concerns from Menin over whether major AI developers have clearly defined financial protections for extreme-loss scenarios.
The questioning extended beyond insurance to the probability of a catastrophic AI event. Menin asked representatives to assign a numerical probability to such a scenario, but none provided a specific figure. An OpenAI representative argued that whether the probability was 1%, 10% or 20%, none of those possibilities would be acceptable.
The hearing also focused on legal responsibility if an AI system were to cause significant harm. Questions addressed potential scenarios involving financial losses, disclosure of sensitive information, physical injury or loss of life, raising broader questions over where liability would ultimately rest when the actions of increasingly capable AI systems result in foreseeable damage.
The issue comes as the New York City Council considers a package of proposed AI regulations. Measures under consideration include independent validation of AI models, mechanisms that could allow systems to be shut down, financial protections for whistleblowers and the possibility of legal claims against AI companies when foreseeable harm occurs.
The discussion highlights a growing distinction between assessing the technical risks of advanced AI and determining who bears responsibility when those risks translate into real-world losses. Insurance coverage could provide one financial mechanism for absorbing damages, while liability rules would determine how responsibility is allocated between developers and other parties involved in deploying AI systems.
Several former researchers from OpenAI, Anthropic and Google DeepMind also warned during the hearing that rapid development of increasingly capable models without sufficient oversight could create risks of humans losing control over such systems.
For AI companies, the questions raised by the hearing extend beyond model performance and safety measures. The ability to establish clear responsibility, financial coverage and effective intervention mechanisms could become increasingly important as AI systems are deployed in settings where failures could carry substantial consequences.
The regulatory debate will therefore remain focused not only on how powerful AI models become, but also on whether independent oversight, shutdown mechanisms and clearly defined liability frameworks can keep pace with their development.
