Anthropic and OpenAI lobby for regulation to stifle rivals
- Anthropic uses lobbying for strict AI safety rules to block smaller competitors.
- OpenAI and Anthropic are consolidating power via enterprise and consumer markets.
- The race is shifting from model development to regulatory capture and profitable APIs.
Anthropic’s rise isn’t just about better models - it’s about using Washington to build a wall. On All-In, David Sacks accused the AI lab of lobbying for a government permissioning regime that would require approvals for new models or chip sales. This creates a regulatory moat that startups can’t cross, securing market dominance for incumbents.
Anthropic’s technical strategy is the perfect cover. It doubled down on coding as a path to recursive self-improvement, a bet that has captured enterprise IT budgets and reportedly added $6 billion to its annual run rate. Its new Claude Mythos model represents a “step change” in reasoning and cyber capabilities, as confirmed on The AI Daily Brief.
David Sacks, All-In with Chamath, Jason, Sacks & Friedberg:
- Anthropic is sort of the most AGI-pilled of all the frontier labs.
- They made this bet on coding as their way to get to recursive self-improvement.
OpenAI is pursuing a parallel form of capture, but through the market. It shelved experimental features like an ‘adult mode’ to focus on enterprise sales and coding tools, consolidating around profitable revenue streams. As The AI Daily Brief reported, both companies are now in a liquidity race, with rumors of Anthropic targeting an IPO as early as October.
The competition is bifurcating. Chamath Palihapitiya noted on All-In that OpenAI’s revenue is three-quarters consumer subscriptions, while Anthropic’s is almost the exact opposite - heavily weighted toward the developer API market. They own different territories: one has the user, the other owns the workflow.
This corporate maneuvering coincides with a fundamental shift in the AI infrastructure layer. CoreWeave CEO Michael Intrator, also on All-In, said demand is decisively moving from training to inference, which he called the “monetization of the investment.” He dismissed fears of rapid GPU obsolescence as “nonsense,” noting clients sign five-year contracts. The compute is becoming a stable, long-term utility, and the companies that control its rules and its pipes are locking in their advantage.
The frontier AI race is over. The moat-building phase has begun.