OpenAI has urged British lawmakers to move now on AI legislation, saying the government should “take advantage of this political window which is clearly opening up” — a remarkable plea from the company behind ChatGPT that lands alongside a cross-party parliamentary committee’s warning that no country on Earth currently regulates AI fit for purpose.
🔍 THE BOTTOM LINE: The world’s most valuable AI lab is now lobbying governments to regulate it, and the UK’s own human rights committee says “nowhere in the world” has a regulatory approach that works. Yet the sharpest fact in the whole story is buried at the bottom: the UK’s AI Security Institute can test frontier models, but it cannot compel any lab — and Anthropic reportedly declined to send its latest model in at all.
What OpenAI actually asked for
In comments to The Guardian on Monday 14 September, Tom Duff Gordon, OpenAI’s head of policy in Europe, said: “We support stronger UK rules for the handful of companies, including OpenAI, developing the most powerful AI systems.”
The qualifier matters. OpenAI wants legislation “narrowly focused on the most advanced AI labs and not startups working on less powerful systems” — a regulatory perimeter drawn around a handful of companies, with itself on the inside of it. In a separate interview with Politico, Duff Gordon added that ministers should legislate while fear of the technology is politically salient, calling it a window that is “clearly opening up”.
The committee’s verdict: nowhere is fit for purpose
OpenAI’s intervention landed alongside a report from the UK’s cross-party joint committee on human rights — MPs and Lords — which concluded that a new regulatory framework is needed, including an independent oversight body and legislation to protect the public.
The committee’s list of AI harms reads like a roll call of problems already live in production: public face-scanning, explicit deepfakes, and systems flagging workers for disciplinary action “without reasonable cause”. Its chair, Labour MP Alex Sobel, said the world was unprepared for the “potentially dire” consequences of a rapidly developing technology, and that “nowhere in the world, including the UK, has a current legislative and regulatory approach to AI that is fit for purpose”.
His prescription is blunter than anything in the government’s current posture: a single AI regulator to set policy, monitor performance and enforce, plus mandatory transparency requirements across the AI life cycle and prohibitions on uses “incompatible with human rights”.
The government’s response was the familiar middle distance. A spokesperson said the UK takes the risks from advanced AI “extremely seriously” and that any future measures would be “evidence-based and focused on the risks we need to address” — which is what the UK has said, in nearly identical words, for two years.
The weekend that produced the window
None of this is happening in a vacuum. Dario Amodei’s Saturday call for the leading labs to “pace the frontier” — including a request that the US government grant an antitrust waiver so rivals could legally coordinate — drew support within hours from Sam Altman, Google DeepMind’s Demis Hassabis and Elon Musk, as we covered in the weekend pause consensus.
Altman’s own position, laid out Sunday night, was subtly different from OpenAI’s UK stance: he would “welcome a federal framework that sets consistent safety requirements for frontier AI”, but insisted there is no need “to wait for an anti-trust exemption or legislation to begin the work” — and that where OpenAI needs government is international coordination, not domestic rules. The full arc, from Amodei’s waiver request to Altman’s counteroffer, is in our coverage of Altman’s response.
So when OpenAI tells UK ministers to seize a “political window”, it is asking for something specific: rules that bind everyone, written while the labs’ safety warnings are still front-page news — and written, ideally, in a form the labs have helped scope.
The oversight gap the report exposes
Here is the part that should follow New Zealand’s debate too. The UK already has an AI Security Institute (AISI) that subjects advanced models to safety tests — but it has no power to compel AI developers to submit their technology. Last week it was reported that Anthropic declined to send its latest model to AISI for testing at all.
That means the UK’s much-admired testing body operates on consent. A lab can simply not show up. We have covered what AISI testing does find when models are submitted — including China’s Kimi K3 being assessed for cyber offence capability — but the regime itself remains voluntary from top to bottom. It is the same structural hole the government left when it rejected legislating an emergency off-switch for frontier models earlier this month: the UK has safety infrastructure, not safety authority.
The committee’s answer — a single regulator with statutory powers across the whole AI supply chain — is aimed directly at that gap. Whether any government moves while the window is open is, as Sobel implies, the actual test.
❓ FAQ
What exactly did OpenAI ask the UK government to do? To legislate quickly while public concern about AI safety is high. Tom Duff Gordon said the UK should “take advantage of this political window which is clearly opening up”, with rules focused on the most advanced labs rather than startups, according to The Guardian.
Why is OpenAI asking for regulation of itself? The stated reason is that consistent rules create certainty for the companies building frontier systems. The political context: the labs have spent the past week urging coordinated slowdowns after Anthropic’s Amodei called for the frontier to be “paced” — see the weekend pause consensus.
What did the UK human rights committee recommend? A new AI bill with steeper obligations for higher-risk systems, bans on AI uses incompatible with human rights, mandatory transparency across the technology’s life cycle, and a single independent AI regulator with enforcement powers.
Did the UK government agree? No. It said it takes AI risk “extremely seriously” and that any future measures would be “evidence-based” — language it has used since before this committee existed. It has already rejected one binding mechanism this month, the proposed AI kill switch.
What does this mean for New Zealand? New Zealand has no binding AI legislation at all and no equivalent of AISI — its two-year-old AI strategy named a skills gap and has moved on little since. If the UK adopts a single AI regulator, NZ firms selling into UK markets would inherit those obligations regardless of what Wellington does.
🔍 THE BOTTOM LINE
The “political window” framing is the most honest thing an AI executive has said about regulation this year: rules get written when fear peaks, and fear peaks are brief. But the same report shows why the window may not matter — the UK’s testing institute cannot compel a submission, Anthropic reportedly tested that freedom last week, and the government’s response to a committee calling the entire global regulatory landscape unfit was to promise evidence-based caution. The regulated are asking for a regulator. Whether they get one with teeth is a different question.
📰 Sources
- The Guardian — “OpenAI urges UK lawmakers to rein in technology amid growing safety fears” (Dan Milmo and Robert Booth, 14 September 2026)
- OpenAI comments via Tom Duff Gordon, head of policy in Europe (The Guardian and Politico, 14 September 2026)
- UK Joint Committee on Human Rights report on AI and human rights, as reported by The Guardian (14 September 2026)
