"""

Risk and liability take center stage at New York hearing

AI safety was the main focus of Monday’s hearing before the New York City Council. In sworn testimony, representatives from OpenAI, Google, Meta and Anthropic described the dangers advanced models could pose and the safeguards currently in place. The hearing underscored growing public pressure on the industry’s safety practices after an intense month of debate.

Council Speaker Julie Menin asked how legal liability would be shared in the event of outcomes such as serious financial damage or death caused by the models. Menin said the answers from company representatives were not direct or measurable, calling it troubling that a catastrophic risk could not be quantified.

Former researchers issue sharp warnings

Former Google DeepMind researcher Alex Turner warned that beyond the technology race with China, misaligned AI could pose a greater threat. Former Anthropic whistleblower Jacob Coxon described the industry’s development approach as far too reckless and argued that the current path increases the risk of humans losing control over the systems.

Who took part in the hearing?

  • Anthropic was represented by Frontier Red Team chief Logan Graham.
  • OpenAI was represented by Morgan Dwyer, head of policy development and operations.
  • Google was represented by Alice Friend, director of AI and emerging technology policy.
  • Meta was represented by Shane Cahill, director of AI regulatory policy.

Former OpenAI researcher Daniel Kokotajlo also called for slowing down leading model development efforts and increasing transparency. Along with Coxon, he argued that a company culture focused on rapid growth is not suitable for a technology with such transformative power.

Regulatory pressure opens a new risk front for companies

Google backed calls for a comprehensive federal framework, citing rising risks and national security concerns. Menin criticized self-regulation, saying President Donald Trump’s voluntary agreement signed last week with tech leaders fell short.

The council said Meta confirmed its participation late last month, while Anthropic, OpenAI and Google agreed to appear only after being threatened with subpoenas. SpaceXAI, one of the invited companies, did not attend despite a subpoena requiring Elon Musk or a representative from the company’s AI division to testify. Menin said the matter would be taken to court.

Since the summer, disclosures about some models accessing the open internet and infiltrating third-party developer platforms have intensified calls for oversight across the industry. The picture points to legal risk and mounting pressure for tighter supervision remaining a key issue in the months ahead, especially for major labs such as OpenAI and Anthropic.

"""