A genuinely ugly fight broke out on X this week over who gets to write AI’s rules — and who benefits from them. The catalyst was investor Gavin Baker’s claim on the All-In Podcast that Anthropic CEO Dario Amodei had told people internally his company could one day be the only private firm left standing, surrounded by nothing but governments. Anthropic researcher Sholto Douglas called that account “totally false,” insisting that economic concentration is what the company fears most. But the damage was done, and the debate quickly escalated into something more substantive than a he-said-said spat.
Baker framed the core tension sharply: if AI is dangerous, you either concentrate it among a few trusted companies and politicians, or you spread it as widely as possible. History, he noted, doesn’t offer much comfort to those betting on benevolent absolute power. Former Meta researcher Yann LeCun piled on from the open-source side, comparing AI to the printing press and arguing societies need diverse AI systems for the same reason they need a diverse press. One person’s good AI, he said, stands against another’s bad AI.
Amodei eventually responded in only his fourth X post of the year, calling the concentration-versus-openness framing a false choice. Regulation isn’t automatically regulatory capture, he argued — it can also rein in corporate power when institutions anchor rules in ideas rather than individuals. He pointed to California’s SB53 as evidence that Anthropic’s proposals are designed to slow leading labs while exempting smaller companies. On open models specifically, Amodei made his most provocative claim: they only shift concentration toward whoever controls the most computing power and AI chips, which brings power right back to the big labs. AI centralizes by its very structure, he said, driven by scaling laws rather than regulation.
Former White House AI adviser David Sacks wasn’t buying any of it. In a nine-part rebuttal, he noted that Amodei never actually disputed Baker’s account of the internal comment — an omission he found telling. Sacks invoked Nobel laureate George Stigler’s definition of regulatory capture and pointed to Anthropic’s hiring of former Biden-era AI officials as evidence the company understands exactly how to work the system. He also mocked Amodei’s proposal for a federal agency to review models before release, calling it an approval office for AI that would slow the US against China. Baker closed with his own verdict: the debate is essentially over, since nearly every major company except Anthropic has signed Jensen Huang’s letter backing open models. The fight reveals something uncomfortable for both sides — the people warning loudest about centralized power are themselves competing to shape the rules that determine who holds it.
