Anthropic CEO Dario Amodei argued that AI tends to centralize around compute power and infrastructure, even when models are open, The Decoder reports. His comments came amid a public fight over AI regulation involving investors, policy figures, and researchers.
Critics including Gavin Baker, David Sacks, and Yann LeCun accused Amodei of using fear about AI risks to support rules that could benefit Anthropic. Amodei countered that regulation can also limit corporate power, and that open models alone may simply shift influence toward whoever controls the chips.
The dispute matters because it separates two questions that are often treated as one. Open access can make model technology more widely available, but training, serving, and competing at the frontier still require large amounts of computing capacity.
The practical issue for policymakers is whether openness is enough to keep the market competitive and safe. The argument suggests that rules aimed only at model weights may miss the infrastructure layer where much of AI power is concentrated.