TLDR:CoinFund’s Brukhman says Anthropic’s export control compliance confirmed AI models are the biggest target for government control.Distributed GPU compute already exists to train frontier AI models, but new algorithms are needed to make decentralized use viable.Teams like Gensyn, Prime Intellect, and Pluralis are proving that distributed AI training is feasible and cost-competitive.Pluralis proposes tokenizing AI model weights among participants to create a sustainable business model for decentralized AI.Decentralized AI could serve as a critical counterweight to growing government control over artificial intelligence models. CoinFund founder Jake Brukhman made this argument following Anthropic’s compliance with U.S. AI export controls. He warned that centralized AI development poses increasing risks of unilateral censorship. Brukhman pointed to distributed GPU networks and open decentralized systems as viable alternatives. His comments have reignited debate about the future governance of frontier AI models.Brukhman Links Anthropic’s Export Control Move to Centralization RiskJake Brukhman has been tracking the intersection of AI and decentralized networks since 2020. He argues that AI models are, by nature, a centralizing force in the technology landscape. Anthropic’s compliance with U.S. export controls, he says, confirmed what many in the space already suspected.In a post on X, Brukhman wrote that the development became “market fact” overnight. He framed it as a turning point for how the industry should think about AI governance. His concern centers on the risk that AI could fall under unilateral state control.Unlike many investors in crypto, I did not pivot to AI in the last few years. However, since 2020, I built some of the deepest understanding in this industry on the intersection of AI and decentralized networks (crypto, web3).From the start, it was very clear that AI models are…— Jake Brukhman (@jbrukh) June 13, 2026Brukhman noted that commodity GPU compute already exists in sufficient quantity to support frontier model training. The barrier, he argues, is not availability of hardware but rather the algorithms needed to use it efficiently. Several research teams are now addressing that exact problem.He cited Gensyn, Prime Intellect, Bagel, Pluralis, Nous Research, Macrocosmos, and Covenant AI as teams working on distributed training. Their research, he said, was once widely dismissed as impossible. Today, it shows that distributed training is not only feasible but can be cost-competitive with centralized approaches.Tokenized AI Models Emerge as a Potential Business ModelOpen source AI models have gained wide adoption, yet they face a persistent challenge around economic sustainability. Without a viable business model, open models struggle to attract long-term investment and development resources. Brukhman acknowledged this gap directly in his commentary.Among the teams he cited, only Pluralis has proposed a concrete solution to this problem. The approach involves splitting model weights among network participants through a tokenized structure. This creates a financial incentive for contributors while maintaining decentralized control of the model.The tokenized model structure means no single entity holds full control over the AI system. Participants share ownership of the weights, making unilateral censorship or control significantly harder to execute. Brukhman sees this as a foundational step toward economically sustainable decentralized AI.Brukhman closed his argument with a direct question to the broader industry. He asked whether AI would become fully centralized under government oversight or whether public, open networks would prevail. The answer, he suggested, depends on whether the industry acts on the momentum now building in decentralized AI research.The post CoinFund Founder Says Decentralized AI Can Counter Government Control of AI Models appeared first on Blockonomi.