When world leaders gathered last General Assembly, artificial intelligence was high on their agenda. And rightfully so. Climate change has taught us, at enormous cost, what happens when governments respond too slowly to a technology that transforms the world.
Fossil-fueled industrialization delivered extraordinary improvements in human welfare while destabilizing the climate. AI presents a similar dilemma: tools that can accelerate medical research and expand access to knowledge can also disrupt livelihoods and threaten human control over consequential decisions.
Both concentrate economic power in companies whose incentives reward expansion and can make them resistant to limits.
But there is a familiar solution in the way we addressed climate change: establish the science; bring every nation to the table; and turn shared interests into obligations that governments can enforce.
The United States and China are central to both challenges. Their participation is essential, but even an agreement between them would be insufficient. AI systems, their benefits and their harms, cross borders. Countries that develop little of the technology will still experience its effects on employment, public services and life in every imaginable way.
They deserve a voice in the rules.
The U.N. has already taken an important first step. In August 2025, the General Assembly established an Independent International Scientific Panel on AI and a Global Dialogue on AI Governance. The panel issued a preliminary report in July, and the first dialogue met in Geneva. Now it needs a path from scientific discussion to binding agreements.
There is precedent for this. The Intergovernmental Panel on Climate Change (IPCC), established in 1988, helped lay the scientific groundwork for the U.N. Framework Convention on Climate Change that was adopted in 1992. Those assessments gave governments a common reference point for the recurring negotiations that created a mechanism for collective action.
The AI panel needs experts from every nation, as we did with the IPCC, and genuine independence from both governments and technology companies. Conflicts of interest should be disclosed. Researchers should have secure access to the evidence they need to test industry claims. And the panel should report plainly on what AI can actually do, which risks are supported by evidence, and whether proposed interventions work. When the technology changes, it should be able to issue rapid updates – and distinguish risks we know about from uncertain worst-case scenarios.
As with the annual U.N. climate summits, known as COP, participating nations should establish annual meetings of all participating nations to negotiate safeguards and review implementation based on the science and the actual threats. The real work cannot stop when the conference ends. Technical standards may need updating through an agreed, accelerated process when serious new risks emerge. Like climate change, AI cannot wait for a diplomatic calendar to catch up.
Replicate IPCC successes
The goal should be enforceable treaties and protocols, implemented through national laws. Countries should require independent testing of the most consequential systems, reporting of serious incidents and access to regulators to ensure companies are in compliance. Regulators need authority to impose penalties and restrict dangerous deployments, subject to due process. International review should identify failures and require corrective action. We cannot rely on companies to police themselves.
Countries should not all carry the same burden. Each country’s responsibilities should reflect their contribution to the risks, its capacity to respond, and the impacts on its people – much like the impacts of climate change.
Nations hosting the largest developers should shoulder greater obligations for oversight and funding for countries with fewer resources that need technical assistance and affordable access to beneficial AI. Rules should be proportionate. We should not build a regulatory system that shields the biggest AI companies of today from tomorrow’s competitors.
Critics will object that more than three decades of climate negotiations have left us with worsening fires, floods and heat. They will argue that the U.N. climate process has failed and is far too slow for AI.
They have a point. Climate diplomacy has too often tolerated delay, weak commitments and inadequate implementation. But we can also take note of remarkable successes inspired by those meetings and targets. California just celebrated the 20th anniversary of its Global Warming Solutions Act of 2006, which cut emissions 22% in that time, while growing its economy 38%.
But progress, or lack thereof, must also account for sustained obstruction by parts of the fossil fuel industry and their political allies.
And it must recognize what cooperation has achieved. UNEP’s 2025 assessment reports that warming projections have declined from as much as 3.5 degrees Celsius around the adoption of the Paris Agreement to 2.8 degrees under current policies. That is still dangerous territory, but the numbers show that the trajectory can change.
Of course, diplomacy alone did not make that happen. National policies, public investment, and innovators making clean energy cheaper have all contributed. The U.N. process has helped provide common goals, scrutiny and continuity for those efforts. Without that cooperation, the damage would likely be worse. For AI, we should keep what worked and fix what did not – especially its slow speed and weak accountability.
AI has one advantage that climate advocates lacked for decades. Some of its leading developers openly acknowledge grave risks and call for oversight. That is a striking contrast with the resistance climate policy faced from much of the fossil fuel industry. On Sept. 21, OpenAI called for global technical standards. Earlier this month, Anthropic’s chief executive, Dario Amodei, advocated regulation covering all U.S. frontier AI companies, including those unwilling to cooperate.
We do not have to start from scratch.
We know what delay has cost us on climate change. We also understand the value of getting the rules right. World leaders talked about AI in New York. Now they should give the U.N. the authority to act.
Terry Tamminen is a former CalEPA Secretary under Gov. Arnold Schwarzenegger and current president and chief executive of AltaSea at the Port of Los Angeles.
The post OpEd: World Needs U.N. Convention on AI appeared first on Los Angeles Business Journal.
