There is a particular silence that settles over a governance forum when a proposal arrives that seems, on its surface, to be about safety but reads, upon closer inspection, like a blueprint for market consolidation. I have been listening to that silence for years, from the whitepaper audits of 2017 to the DAO treasury designs of 2024. So when David Sacks, the former PayPal COO and a prominent voice in Silicon Valley's libertarian wing, publicly accused Anthropic of "regulatory capture," I did not hear a scandal. I heard a confession about how the game is actually played.
The accusation, reported by Crypto Briefing, is a blunt instrument. Sacks alleges that Anthropic, the maker of the Claude model family, is leveraging its "safety-first" positioning to lobby for stringent AI regulations that would disproportionately burden open-source competitors. The logic is as old as commerce itself: if you cannot beat the competition on merit, change the rules of the arena. This is not a technical debate about model alignment or hallucination rates. It is a political battle for the architecture of the AI industry's future, and the weapon of choice is the very language of protection we once reserved for the vulnerable.
To understand the gravity of this, we must strip away the marketing. Anthropic has built its brand on a narrative of cautious, ethical AI development. They have positioned themselves as the responsible adult in a room full of reckless children. This is a powerful story, and it has attracted billions in funding from Microsoft, Google, and Salesforce. These are not investors who are naive about the power of regulatory moats. They understand that compliance is a fixed cost that scales beautifully. A million-dollar audit is a rounding error for a hyperscaler, but it is a death sentence for a small startup trying to fine-tune an open-source model in a garage.
Sacks's intervention, filtered through my own experience auditing governance structures, points to a systemic truth: the most effective moats are not built with code, but with clauses in legislation. The term "regulatory capture" describes a process where regulated entities gain undue influence over the regulator, effectively writing rules that serve their own interests. In the context of AI, this translates to a scenario where a dominant closed-source player advocates for safety standards that are so expensive and complex that only they can reasonably comply. The narrative of protecting humanity from rogue AI becomes a convenient shield for protecting market share from disruptive innovation.
This is where my contrarian lens focuses. The open-source community, in its purest form, believes in a meritocracy of code. They argue that transparency is the ultimate safety mechanism—that a million eyes looking at the weights and the training data will find flaws faster than any internal red team. There is profound truth in this. The ability to audit, replicate, and fork is a form of distributed accountability that no closed system can match. Yet, the open-source world is not a monolith of virtue. It is also a landscape where powerful actors, like Meta with its Llama models, use "openness" as a strategic weapon to commoditize their competitors' offerings. The accusation of regulatory capture, therefore, is not a simple story of good versus evil. It is a story of two oligopolistic tendencies fighting for control.
My skepticism is my shield here. I refuse to accept Sacks's framing at face value, just as I refuse to accept Anthropic's safety narrative without question. Sacks is not a neutral observer. He is a venture capitalist with deep ties to the crypto and open-source worlds, and his portfolio likely benefits from a regulatory landscape that is permissive towards decentralized, open projects. His accusation is a political move, designed to rally a specific constituency. But that does not make it false. In fact, his position highlights a fundamental tension: we are asking governments to regulate a technology that is still in its infancy, and we are doing so while the largest players are actively trying to shape those regulations to their advantage.
The real risk here is not that Anthropic succeeds in crushing open source. The risk is that the debate becomes so polarized that we end up with a bifurcated ecosystem—one where "safe" AI is only accessible through a few corporate gatekeepers, and "open" AI is relegated to a legal gray zone, forcing innovation underground or overseas. This is the fragmentation scenario I fear most. I have seen it happen in DAOs, where well-intentioned governance rules are gamed by whales until the community splinters into factions. The ledger remembers the transactions, but the community must forgive the transgressions to survive.
The alpha in this situation hides in the boredom of due diligence. We are not watching a technical breakthrough; we are watching the slow, deliberate construction of a regulatory apparatus. The questions we should be asking are not about which model is smarter, but about who gets to define the standards. Will the EU AI Act, with its nuanced tiers, create a safe harbor for open-source innovation? Or will it, under lobbying pressure, become a tool for incumbent protection? The answers will determine the value chain of the next decade, not just for AI, but for the entire digital economy that builds upon it.
Truth is coded in transparency, not promises. Anthropic's promise of safety is meaningless if it is used as a cudgel to prevent others from building. Sacks's promise of freedom is equally hollow if it simply allows the most powerful open-source actors to operate without any accountability. The path forward is not to choose a side, but to design a governance framework that is resistant to capture by either. This means demanding radical transparency from regulators, requiring disclosure of lobbying activities, and creating technical standards that are truly neutral. We must build the shield of skepticism and wield the sword of empathy, understanding that the fear of AI is real, but so is the fear of centralized control.
As we move deeper into this bull market of AI hype, the technical flaws are not in the code—they are in the incentives. The next great protocol will not be the one with the best benchmark scores, but the one that can navigate this political minefield without losing its soul. The question we must carry forward is not whether Anthropic is guilty of regulatory capture. The question is whether we, as a community of builders and thinkers, are capable of creating a system where the loudest voice in the room is not the one with the most money, but the one with the most compelling argument for the common good. The silence between the code lines is getting louder, and it is asking us to listen.
