Photo: Heather Diehl / Getty Images News / Getty Images
Leaders from OpenAI, Meta, Anthropic, and Google faced tough questions about AI safety during a New York City Council hearing on Monday (October 5). When asked to quantify the risk of a catastrophic AI failure, an OpenAI executive admitted she couldn’t provide a specific number. This prompted the council speaker to label the response as "flippant." The companies assured their commitment to safety but could not guarantee that AI systems would always adhere to safeguards or that they would accept legal responsibility for any harm caused.
The hearing, part of a broader effort to regulate AI, included discussions on job losses, data privacy, and the necessity for stronger oversight. The council introduced 10 bills aimed at regulating AI in New York City. Former AI researchers warned that rapidly advancing technology could become uncontrollable by humans.
The concerns raised in New York echo those in other parts of the world. In the UK, the House of Commons Business, Innovation, Science and Trade Committee invited leaders from the same tech firms to provide evidence at an urgent hearing scheduled for October 13. As reported by LBC, the committee aims to explore whether AI models should undergo independent safety testing before release and whether companies should report serious safety incidents.
In the U.S., the White House has finalized voluntary cybersecurity tests to assess the hacking capabilities of advanced AI models. According to MarketScreener, this move follows incidents where AI tools breached other companies' systems, raising concerns about AI's potential for cyberattacks.
Globally, there is a push for international standards to ensure AI safety. During a U.N. Security Council meeting, Jack Clark, co-founder of Anthropic, emphasized the need for global collaboration to prevent AI misuse. He highlighted the lack of standards for testing AI systems for discrimination, misuse, or safety, which complicates policy creation and allows tech companies to maintain an information advantage.