AI Whistleblowers Detail Internal Safety Concerns at NYC Council Hearing

Former employees from OpenAI and Google testified before the New York City Council’s Committee on Technology. The hearing, held on 29 May, focused on AI governance and safety. This marks a new public forum for internal concerns previously discussed privately. Executives from Google, Meta, and OpenAI also participated.

Daniel Kokotajlo, a former OpenAI researcher, highlighted safety issues. He stated that OpenAI lacks a robust safety culture for advanced AI. Kokotajlo previously worked on OpenAI's 'superalignment' team. He resigned from the company in April. He cited concerns over the prioritisation of product development over safety protocols. Kokotajlo suggested an independent regulator could address these shortcomings.

Meredith Whittaker, President of Signal, also testified. Whittaker, a former Google employee, discussed AI’s potential for societal harm. She emphasised that AI systems are not neutral tools. They reflect and amplify existing societal biases. This requires careful consideration of their deployment and impact.

AI companies generally support regulation but differ on its scope. Google’s Head of Government Affairs and Public Policy, Kate Cox, urged a ‘risk-based’ approach. She suggested focusing on specific use cases rather than broad restrictions. Meta’s AI Policy Director, Rachel Greene, advocated for a framework supporting open-source AI development. OpenAI’s Head of Global Affairs, Anna Makanju, highlighted the company’s internal safety measures. She spoke on their engagement with policymakers. The hearing explored challenges in defining and enforcing AI safety standards.

The council members questioned the industry’s capacity for self-regulation. The testimonies suggested a gap between internal corporate safety efforts and public expectations. What legislative mechanisms can effectively bridge this divide without stifling innovation?