Former AI Researchers Testify to NYC Council That Humanity Is More Likely Than Not to Lose Control of AI

Former researchers from OpenAI, Anthropic, and Google DeepMind testified before the New York City Council, warning that humanity is more likely than not to lose control of advanced AI. Jacob Coxon, a former researcher at both OpenAI and Anthropic who resigned from Anthropic in September, stated that the current trajectory could end in human extinction. He described AI companies as "gambling" with human lives.

Testimony Highlights

• Jacob Coxon (ex-OpenAI and Anthropic): Asserted that AI companies are "gambling" with human lives and that losing control is the probable outcome on the current path.
• Daniel Kokotajlo (ex-OpenAI): Testified that the field resembles psychology more than engineering because AI systems are "trained or grown" rather than designed. He noted that the ability to notice misalignment problems is already poor and expected to worsen.
• Alex Turner (ex-Google DeepMind): Left the company in June after it signed a Pentagon deal he opposed. He was subpoenaed to testify, as was Kokotajlo.

Legislative Context

The hearing was the first Committee of the Whole session for the City Council since 2022, convened to evaluate a package of AI bills proposed by Speaker Julie Menin. The testimony aimed to inform legislative decisions on AI safety and regulation.