Superintelligence is no longer a fringe concern in Washington, Brussels, or Beijing. The question is whether governments can move faster than the technology they are supposed to govern.
In November 2023, 28 countries signed the Bletchley Declaration, the first international agreement to acknowledge that advanced AI poses risks that are "potentially catastrophic." The signatories included the United States, China, the United Kingdom, the European Union, and nations across Asia, Africa, and South America. This was not a fringe document. It was signed by heads of state and senior ministers.
The UK hosted the first AI Safety Summit at Bletchley Park. The second was held in Seoul. A third followed in Paris. An international network of AI Safety Institutes now exists, with institutes in the United States, United Kingdom, and European Union sharing research and coordinating on risk frameworks. None of this existed four years ago.
In the United States Senate, AI risk has become one of the rare genuinely bipartisan issues. Chuck Schumer convened the AI Insight Forums in 2023. Josh Hawley has called superintelligence "the most dangerous technology in human history." Marco Rubio has introduced legislation targeting AI risk. The concern cuts across party lines because the threat does not respect ideology.
The European Union passed the AI Act in 2024, the world's first comprehensive legal framework specifically regulating AI systems by risk level. The most capable frontier models face the strictest requirements. It is imperfect legislation. It is also binding law, enforced by regulators with the authority to issue substantial fines.
China has issued its own AI governance regulations, including requirements for algorithmic transparency and content labeling. China participated in the Bletchley Summit and the Seoul Summit. The picture of China as unwilling to engage on AI governance is, as a matter of fact, wrong.
The political situation in 2026 looks nothing like the political situation in 2020. That shift happened because people pushed for it. It can continue to shift, and must, before the systems being built make the governance question unanswerable.
28 nations, including the US, UK, EU, and China, sign the first international agreement acknowledging that advanced AI poses risks that are "potentially catastrophic." The AI Safety Summit at Bletchley Park produces a shared framework for frontier AI risk evaluation.
The European Union passes the world's first comprehensive AI regulatory framework, imposing strict requirements on the most capable frontier models. Violations carry fines of up to 3% of global revenue. The law is binding across 27 member states and applies to any AI system deployed within the EU.
The US Senate holds a series of bipartisan hearings on AI risk, with testimony from Sam Altman of OpenAI, leading researchers, and policy experts. The hearings mark the first time AI existential risk is discussed substantively at the highest levels of US legislative oversight.
The US, UK, and EU establish AI Safety Institutes with a mandate to evaluate frontier AI systems for catastrophic risk before deployment. These institutes represent the first formal government infrastructure for superintelligence risk assessment.
OpenAI, Google, Microsoft, Meta, and Amazon spent hundreds of millions on lobbying in Washington in 2023 and 2024. The technology industry is the single largest corporate lobbying force in the United States. It employs former regulators, former legislators, and former White House officials. This is the force that any AI governance effort must overcome.
The lobbying is not random. It is targeted at the specific legislative mechanisms that could constrain superintelligence development: compute thresholds, mandatory safety evaluations, liability frameworks. Every measure that would create real accountability is met with a well-funded campaign to weaken or defeat it.
AI capability has advanced from GPT-2 to systems that outperform most humans on most cognitive benchmarks in under six years. Major legislation typically takes four to eight years from introduction to passage. The technology does not wait for governance to catch up.
This mismatch argues for treating legislative urgency as a first-order priority, for funding the policy work that compresses timelines, and for building the public pressure that forces faster action. The speed mismatch is a political problem, not a physical law.
Safety regulations typically follow disasters, not precede them. Seatbelt laws came after decades of traffic deaths. Financial regulation tightened after crises. The political mechanism for precautionary regulation is poorly developed in most democracies.
For superintelligence, the disaster-first model is not acceptable. A misaligned superintelligence does not offer a corrective period. This is why precautionary governance, modelled on nuclear non-proliferation rather than post-crisis regulation, is the right framework. It has worked before. It requires building the political infrastructure to make it work again.
The tobacco industry said it was impossible to restrict cigarette advertising. The auto industry said seatbelt mandates would destroy car sales. The chemical industry said banning CFCs would end modern manufacturing. In each case, the claim of impossibility was a lobbying position, not a fact. In each case, the regulation happened anyway.
The Overton window on AI risk has shifted faster than any comparable technology issue in history. In 2020, superintelligence risk was a topic discussed primarily in academic papers and online forums. In 2026, it is on the agenda of every G7 government. Senior heads of state use the phrase "existential threat" in public addresses. That shift was not inevitable. It was the result of sustained advocacy by researchers, journalists, and citizens who refused to let the issue be dismissed.
The political coalition for AI safety is forming. It includes conservatives who worry about national security and the concentration of power in a handful of corporations. It includes progressives who worry about the absence of democratic oversight over the most consequential technology decisions in history. It includes faith communities, scientists, artists, and parents. This is the beginning of a majority.
International momentum is real. The Bletchley process, the AI Safety Institute network, and the emerging UN governance discussions represent the fastest construction of international AI policy infrastructure in any technology domain. The pieces are assembling. The question is whether they will assemble fast enough, and that question depends on whether enough people push.
Join us and help build the pressure that makes inaction politically untenable.