Former OpenAI and Anthropic researcher Jacob Coxon has urged greater transparency from leading AI companies, warning that developers do not yet have complete control over the behaviour of their most advanced models.
Coxon made the remarks while testifying before the New York City Council, shortly after resigning from Anthropic. He called on AI companies to provide greater access to information for the public and independent experts as the development of increasingly powerful systems continues.
The whistleblower said he believes companies should be more open about the limitations and risks associated with their models, arguing that independent scrutiny is necessary to assess whether safeguards are working effectively.
Coxon also called for a pause in the development of what he described as frontier AI systems until researchers can establish greater confidence in their safety. “We should slow down frontier AI development until we can be confident that it’s safe,” he told the council.
His testimony comes amid growing debate over how much control AI developers have over increasingly capable models and whether existing safety measures are sufficient to manage unexpected or potentially harmful behaviour.
AI companies have invested heavily in research aimed at improving model reliability, alignment and safety, while governments and regulators in several jurisdictions are developing rules to govern the technology.
Coxon’s comments add to calls from some researchers and former industry employees for stronger independent oversight and greater transparency around how advanced AI systems are trained, evaluated and deployed.
His testimony before the New York City Council is likely to add to the wider debate over the pace of AI development and the responsibilities of companies building increasingly powerful systems, with discussions over regulation and safety continuing.
