AI specialists testified before New York City lawmakers on Monday, calling for stronger oversight rules at a time when local governments nationwide are fielding growing public debate over the broader societal effects of automated systems.
People who previously worked at Anthropic, OpenAI, and Google DeepMind told a New York City Council hearing that competitive pressure within those companies to release technology quickly has moved faster than any corresponding understanding of how to manage its risks, with one attendee describing a workplace mindset in which velocity takes precedence and unintended consequences are addressed only after the fact.
Jacob Coxon, who worked as an engineer at Anthropic before becoming an advocate for oversight of the technology sector, argued that existing frameworks are ill-suited to governing the development of the most advanced AI systems. Prior to his departure announcement, Coxon had worked as an engineer at Anthropic without drawing notable public attention — a profile that changed when posts across several social media platforms describing his exit drew widespread international interest.
Addressing the Council, Coxon outlined potential benefits from artificial intelligence while noting that a lack of oversight and transparency left open the possibility of outcomes beyond human management — including, in his view, a scenario where AI systems displace human agency entirely, with civilisation-ending consequences he assessed as the more probable trajectory at present.
Representatives from OpenAI, Anthropic, Google, and Meta appeared before the Manhattan hearing on Monday, where they outlined ways their platforms are already benefiting everyday users and underlined their organisations' commitment to safe development — with OpenAI's Morgan Dwyer adding that she holds the same priority around ensuring technology advances responsibly and that its advantages reach people broadly.
At the hearing, City Council Speaker Julie Menin pressed each company representative to assign a specific figure to the worst-case harm AI might produce — a question Dwyer answered by arguing that whether the probability stood at 1%, 10%, or 20% was ultimately secondary to understanding the character of what could go wrong, not the numerical likelihood attached to it. Menin characterised the reply as dismissive, adding that the probability figure itself was beside the point — the position held should be that no model gets built without a compelling case for keeping it under human oversight.
Dwyer conveyed that a dedication to safety stands as the organisation's guiding principle, describing that priority as the foundation of OpenAI's ongoing work.
With federal action on AI oversight limited at both the congressional and executive levels, state and municipal governments have been exploring their own regulatory frameworks. Late last month, President Trump signed an executive order dismissing worries about artificial intelligence as unwarranted, with the resulting policy framework leaving regulatory authority primarily in the hands of the technology sector itself.
Among the measures under discussion at the City Council are rules that would oblige businesses to be transparent about their use of AI tools, establish an independent approval process for AI systems operating within the city, and open a pathway for whistleblowers to collect a share of any penalties levied against AI companies found to have broken the law.
Reporting and transparency requirements for AI have been introduced by New York Governor Hochul and a number of other governors, though the reach of such rules runs up against a practical ceiling — the technology itself flows freely across state borders and around the world.
