Former Anthropic researcher Coxon testifies at NYC Council AI hearing

By Splitfeed EditorialLast updated Oct 6, 2026, 22:50 UTC

On October 6, 2026, the New York City Council convened a hearing where executives from Google, Meta, OpenAI, and Anthropic testified under oath about AI safety. Jacob Coxon, a former researcher at both Anthropic and OpenAI who left over safety concerns, appeared at the invitation of Council Speaker Julie Menin, who had actively sought whistleblower testimony as the Council weighs a package of AI safety bills. Coxon called AI firms "reckless," argued they cannot control their models, and advocated for kill switches and stronger regulation. OpenAI separately sent the Council a letter recommending safeguards ahead of the session.

The hearing is notable because municipal governments rarely compel major AI companies to testify under oath. Council Speaker Menin has signaled the Council will pursue legal action against at least one company, SpaceX AI, for defying a subpoena and failing to appear, according to Lora Kolodny. That enforcement posture, combined with the whistleblower framing of Coxon's appearance, raises the stakes beyond a typical advisory session. The core dispute is whether AI companies are engaged in reckless development they cannot govern, or whether regulatory pressure from a city government is an overreach that misunderstands the technology.

The anti-AI camp is using the hearing as evidence of systemic evasion. Jessy Edwards noted that executives could not answer basic questions, including how many staff work on extinction risk, and separately pointed out that OpenAI, Anthropic, Google, and Meta lobbied against the Raise Act for a year. The Staten Island Advance and Carol Roper both described the companies as racing to build technology they do not understand how to control. The middle ground camp is focused on the procedural mechanics of oversight itself. Techpresso noted Coxon's appearance was arranged by Menin as part of a broader legislative package, and Mike Brock suggested that if the leading closed AI labs collapsed, investment would shift to open-weight models and might produce better outcomes overall. The pro-AI camp has not yet responded publicly; typically such voices would argue that existing corporate safety teams and voluntary commitments make municipal regulation redundant or counterproductive.

The clearest near-term signal to watch is whether the NYC Council advances any of its proposed bills into law, and whether its subpoena enforcement action against SpaceX AI succeeds in court, which would set a precedent for how much authority local governments can actually exercise over the industry.