OpenAI chief executive Sam Altman has made AI power concentration one of his central public concerns, arguing that a world in which a single company or government controls artificial intelligence would be ‘very, very bad’, even as he runs one of the few organisations with the capability to build frontier models.
‘I think the fight of the current moment is: Are we going to head to a world of AI authoritarianism or liberty?’ Altman said on the latest episode of the ‘Relentless’ podcast. ‘For so long now, I have felt focused on this singular goal of abundant intelligence and a belief that incredible human prosperity will come from that, as long as we don’t have a weird power concentration and kind of a new authoritarianism.’
The comments came as a live debate over open-source access to frontier AI models intensified in Washington and Silicon Valley.
The letter that put AI power concentration on Washington’s agenda
The immediate backdrop is a joint industry letter titled Open Weights and American AI Leadership, which called on policymakers not to restrict open-weight AI models. Business Insider confirmed that 32 companies and organisations had signed by Friday evening, a broader coalition than initial reports indicated, including Microsoft, Meta, Nvidia, OpenAI, Cisco, GitHub, Cohere, DoorDash, and Palo Alto Networks.
Google was not among the original signatories but added its name on Saturday. Chief executive Sundar Pichai wrote on X: ‘Very happy to support this on behalf of Google.’
The letter also warned policymakers not to conflate distillation (described in the letter as ‘a widely used technique’) with unlawful model extraction, a distinction that has become central to the policy debate.
Anthropic, OpenAI’s most direct competitor, was the only major frontier lab that neither signed the letter nor publicly supported its contents. The company has consistently argued that increasingly capable models require stronger safeguards. In February 2026, Anthropic updated its Responsible Scaling Policy to Version 3.0, describing it as a comprehensive rewrite that introduced Frontier Safety Roadmaps with detailed safety goals and Risk Reports quantifying risk across all deployed models.
Anthropic has also been deepening its engagement in Washington. The New Stack reports that the company spent more on federal lobbying in the first half of 2026 than it did across the whole of 2025.
Kimi K3 and the Chinese model that sharpened the debate
The catalyst for much of this activity was the release of Kimi K3, a 2.8-trillion-parameter Mixture-of-Experts model built by Beijing-based Moonshot AI, which is backed by Alibaba. Moonshot launched Kimi K3 on 16 July 2026 and published the full model weights on 27 July 2026.
The model’s scale puts it in a different category from earlier open-source releases. VentureBeat reports that Kimi K3’s 2.8 trillion total parameters make it roughly 75 percent larger than DeepSeek’s V4 Pro, which has approximately 1.6 trillion parameters. Kimi K3 runs 104 billion active parameters at inference and supports a context window of over one million tokens.
The release prompted some US officials to call for restrictions on access to Chinese models, while investors and startup founders warned that doing so could damage American competitiveness globally. US export controls already restrict which countries and companies can obtain the most advanced chips and computing infrastructure, giving Washington existing leverage over who can build at the frontier.
Altman’s own position in this debate contains a visible tension. OpenAI keeps its models proprietary rather than open-source, meaning competitors, researchers, and governments cannot inspect the architecture of its systems. His stated goal of preventing power concentration applies in theory to his own organisation as much as to anyone else.
‘Every time that humanity has traded off its liberty for safety, it’s been a long-term net loss,’ Altman said. ‘We are going to put this in the hands of people. We’re going to empower them. We are going to let society express its ideas and use this technology in the way they want.’
Both OpenAI and Anthropic are valued near $1 trillion, according to the Times of India, and both have confidentially filed for IPOs that could land in 2026, meaning the question of who controls the most powerful AI systems may soon be one for public shareholders as much as policymakers.