Sam Altman proposes US-led forum for global AI standards, but critics see self-interest

Altman's AI safety proposal: let us win, or everybody loses

A US-led world order would hand a big advantage to an American oligopoly that includes his own company Sam Altman's argument in these pages that it is time for a "global framework" for AI reflects a stark reality. The technology he and his peers are creating has enormous destructive potential that spills beyond national borders. He is also right that the world has faced such challenges before. This new one, though, is harder to fix. Altman's idea sounds a lot like the Basel Committee on Banking Supervision, applied to AI. He suggests a global forum of policymakers and experts that would set standards for AI models, watch to make sure the profit motive doesn't trump safety, and grant compliant countries access to advanced technologies. As with Basel, enforcement and actual rule-setting would be done by individual countries, to avoid seeming undemocratic. This might sound like a bratwurst lobbying for a July 4 barbecue. After all, AI lobbyists have argued furiously, when confronted with US state-level legislation, that red tape chills innovation. In finance, too, banks grumble loudly about Basel, which lumbers them with huge compliance costs and can appear unduly fraught with complexity. But international bank oversight has had unintended consequences that, if applied to the AI space, might not be half bad for Altman and his peers. For one, it has entrenched the largest lenders. Onerous regulation favours those with budgets large enough to navigate it. And to the extent that the oversight body is US-led, as Altman suggests it should be, US interests will inevitably dominate. Another fly in the ointment: China. To be genuinely effective, a global alliance would have to include the People's Republic -- as Basel and the International Atomic Energy Agency both do. It's hard to see Beijing agreeing to brakes on its technological progress. A global accord could therefore turn into an iron curtain with China on the other side -- an idea that may appeal to western AI firms worried their customers will opt for cheap Chinese rivals such as DeepSeek. Basel demonstrates two other problems with global oversight. First, crises are easier to spot in hindsight, even by experts. Basel's priorities and prescriptions have at various times amplified rather than fixed weaknesses in the system, while focusing on the wrong risks. How that might play out in AI is hard to know. Common standards could lead to shared points of failure, or lead everyone to overlook the same vulnerabilities. Second, Basel took a long time to get going, and when it comes to AI risk, time is in short supply. The committee was set up in 1974, but it wasn't until 1988 that it first settled on rules regarding bank capital. The IAEA took almost four years to create from when US President Dwight Eisenhower called for an international body to address nuclear threats. Altman thinks systems with "astonishing power" are just a year or two away. The OpenAI chief is right to warn that if global collaboration doesn't happen, fragmentation will. Countries will impose barriers of their own; these will almost certainly prove ineffective at keeping out bad actors and dangerous technologies. But a US-led world order would hand a huge advantage to an American oligopoly that includes his own company. In a world of unenviable choices, the latter is as good as it gets.

Sam Altman calls for US-led international forum to set global AI standards

Sam Altman calls for US-led international forum to set global AI standards Sam Altman is calling for a US-led international forum to set global safety standards for artificial intelligence, arguing that no single country should be left to dominate the technology. In an op-ed published Wednesday in the Financial Times, the chief executive of OpenAI Group PBC proposed "a US-led international forum that establishes accepted standards, provides expert and impartial analysis of capabilities and risks and makes the technology available to nations and companies that participate and follow the rules." The body would bring together government representatives, independent technical experts and others, Altman wrote. He said it could "serve as a governance mechanism over the labs and guard against the commercial pressures that can lead to unsafe racing." Altman pointed to global aviation safety rules, international financial standards and the International Atomic Energy Agency as models. The IAEA was set up in 1957 to police civilian nuclear energy and did so even as the U.S. and Soviet Union built up their arsenals. Altman's argument is that rival powers have found ways to govern dangerous technologies together. He framed the plan partly as a way to spread the benefits of AI beyond a handful of companies. "Everyone on Earth should benefit from this technology and determine for themselves how best to use it," he wrote. The proposal follows a Group of Seven summit in France at which executives from OpenAI, Anthropic PBC and Google DeepMind met world leaders and discussed setting common standards for advanced AI models. The idea of an international forum was attributed to Altman at those talks. Enforcement is the open question. Unlike aircraft and nuclear enrichment plants, which inspectors can physically examine, frontier models are trained inside data centers with little outside visibility. That opacity makes it far harder to verify whether a lab is following any agreed set of rules or racing ahead in secret. Altman is not alone in pushing for oversight. OpenAI and Anthropic have both previously backed the idea of an international AI watchdog. Anthropic Chief Executive Dario Amodei has argued for rules closer to those of the Federal Aviation Administration, a more prescriptive model than the standards-setting body Altman describes. The pitch also comes as Washington tightens its grip on the industry. OpenAI has agreed to roll out its coming GPT-5.6 models first to a group of government-approved partners. Anthropic was briefly forced to pull its Fable 5 and Mythos 5 models last month after a Commerce Department order restricting foreign access, then restored Fable 5 this week. Whether the forum gains traction may depend less on the labs than on governments. Writing for the Brookings Institution, analysts argued this week that the G7 should accept the industry's offer to help set AI standards, but only on the condition that any resulting agreement is made enforceable rather than voluntary.