MANY MINDEDAI Power Concentration Risk in California Policy
Anthropic CEO Dario Amodei argues AI systems inherently concentrate power through scaling laws, with open models shifting control to entities owning computing infrastructure. He identifies California's SB53 law as a concrete policy action that exempts companies below specific revenue or training-cost thresholds from AI regulatory requirements—a move he claims could allow Anthropic to become the sole private entity managing AI development after regulatory changes. The dispute originated from investor Gavin Baker's podcast claim, where Amodei reportedly stated Anthropic hires former government officials to shape regulations favorably. This creates friction for security: if SB53's thresholds enable concentrated control, critical AI capabilities could become gated to fewer entities. Caveats: The source cites an August 18, 2026 publication date (likely a typo), SB53's exact thresholds are unspecified, and Amodei's claims are contested by multiple parties including Meta researcher Yann LeCun.
Source: The Decoder