Editorial illustration for Anthropic CEO Says AI Centralizes Power, Open Models Shift It to Chip Owners
Anthropic CEO: Open AI Models Shift Power to Chip Makers
Dario Amodei has spent the past week defending himself on X against a charge that his company's regulatory push is really a power grab dressed up as safety policy. The Anthropic CEO argues AI concentrates power by default, and that open-source models don't fix this, they just hand control to whoever owns the most chips. His critics see it differently.
Investor Gavin Baker, former White House adviser David Sacks, and Meta researcher Yann LeCun all say Amodei is using fear about AI risk to lobby for rules that would box out competitors and lock in Anthropic's position. Sacks has gone further, accusing Anthropic of stacking its ranks with former government officials specifically to steer legislation, and warning that an approval agency for AI models, the kind Amodei has floated, would leave the US at a disadvantage against China. The dispute traces back to a claim on the All-In Podcast that Amodei privately predicted Anthropic could end up as the last private AI company standing, with only governments left beside it.
Anthropic has called that account false.
Baker then spelled out the heart of the dispute: If AI is potentially dangerous, there are two basic stances. Either you consider it too dangerous to spread widely and concentrate it among a few companies and politicians, or you consider it too dangerous to concentrate and spread it as far as possible.
Why this matters
For developers and founders, this fight is really about who gets to write the rulebook while claiming to fear it. Amodei's argument that open models simply hand power to chip owners is a real point, GPU access is already concentrated among a handful of cloud giants, and pretending otherwise ignores that supply chain. But Baker's claim about Amodei privately floating a world with Anthropic as the last private lab standing, denied by Sholto Douglas or not, deserves more scrutiny than a one-line rebuttal on X.
We'd want to see the actual regulatory text Anthropic is pushing, not just Amodei's framing of it. If the rules mainly raise compliance costs for smaller labs and open-source projects, that's centralization dressed up as safety, regardless of intent. Anyone building on open weights right now should watch which specific policies Anthropic lobbies for next, not just the philosophical argument about who "naturally" ends up holding power.
The chip-ownership point is worth taking seriously. The messenger's motives are worth taking just as seriously.
Related Reading
Browse all Anthropic news →
Business & Startups Nordic pilot adds Gemini for Education, NotebookLM to boost AI literacy
Business & Startups Kling launches Video O1, all-in-one model with MVL bridge using transformer
Business & Startups DeepSeek Seeks More Capital Weeks After USD 7B Funding Round
Related Topic Trump cracks down on Anthropic after Amazon tip; staff largely foreign
Open Source
Related Topic Anthropic says Claude writes >90% of its code; AI pause button urged
LLMs & Generative AI
From Archives Commerce AI's Fragmentation Problem: Why Lost Context Costs Sales
From Archives Perplexity's Free AI Offer Drew Millions of New Users in India
From Archives Claude Sonnet 5 Launches With User-Adjustable Effort Level
LLMs & Generative AI
From Archives Anthropic Watermarks AI Output, Some Clients Request It
LLMs & Generative AI
Common Questions Answered
What is Dario Amodei's argument about how open-source AI models affect power distribution?
Amodei argues that while open-source models may appear to democratize AI, they actually concentrate power among whoever owns the most chips and computing infrastructure. He contends that open models don't solve the centralization problem—they simply shift control from a few AI companies to a handful of cloud giants and chip owners who control GPU access.
What are the two opposing stances on AI safety mentioned in Gavin Baker's quote?
According to Baker, one stance treats AI as too dangerous to spread widely and therefore advocates concentrating it among a few companies and politicians. The opposing stance considers AI too dangerous to concentrate and instead argues for spreading it as far as possible to prevent any single entity from gaining excessive control.
Why do critics like Gavin Baker and Yann LeCun dispute Anthropic's regulatory approach?
Critics argue that Amodei is using fear about AI risk as justification for regulatory policies that would benefit Anthropic's business interests. They contend that his emphasis on AI dangers is being used to lobby for regulations that concentrate power rather than genuinely addressing safety concerns.
What does the article identify as a real constraint on open AI model distribution?
The article acknowledges that GPU access and computing infrastructure are already concentrated among a handful of cloud giants, making it difficult for open models to achieve true democratization. This supply chain concentration means that even if models are open-source, the ability to actually run and deploy them remains limited to those with access to expensive chip resources.
Further Reading
- Papers with Code - Latest NLP Research - Papers with Code
- Hugging Face Daily Papers - Hugging Face
- ArXiv CS.CL (Computation and Language) - ArXiv
