David Sacks accuses Anthropic's Dario Amodei of trying to create a "DMV for AI." But plenty of industries thrive despite safety regulation.

The AI world spent its weekend arguing on X, and the fight says a lot about where the industry actually stands. It started when investor Gavin Baker, co-hosting the All-In podcast with former Trump AI czar David Sacks, claimed multiple sources told him Anthropic CEO Dario Amodei believes his company "might be the only private company in the world at some point." Baker framed that as proof of a maximalist Anthropic vision in which only it and the U.S. government get to decide who touches powerful AI.

Sacks, no fan of Anthropic to begin with, pounced. He called the alleged remark hubristic and repeated his long-running accusation that Amodei is running a regulatory capture play — stoking fear about AI so Washington locks in rules only Anthropic can comply with. Sacks went further, warning that any kind of oversight body, which he mockingly labeled a "DMV for AI," would create testing backlogs and hand China an edge by constraining U.S. developers while Beijing faces no such limits.

Anthropic pushed back hard. Chief brand and communications officer Sasha de Marigny called Baker's claim "complete and utter nonsense" on X, though Sacks noted Amodei himself never directly denied the specific quote. What Amodei did do was write a lengthy response laying out a more complicated position than either side of the debate wanted to hear. He argued that Silicon Valley libertarians like Baker default to seeing all regulation as capture, while people outside that bubble see it as protecting ordinary users, and that both views oversimplify a question that depends entirely on what the actual rules say.

Amodei said Anthropic tries to design proposals that specifically disadvantage frontier labs while carving out exemptions for smaller competitors — thresholds based on revenue or training spend, or rules aimed only at the most capable models. He conceded AI naturally concentrates economic power because of the sheer compute needed to train and run it, but said that's not the same as claiming only one or two companies will ever exist. Open-weight models, he added, only partly fix that concentration problem since they still need compute to actually run.

On the public's souring mood toward AI, Amodei rejected the idea that his own warnings are to blame, calling it instead "fundamentally a crisis of trust" that no glitzy marketing campaign can fix. The real fix, he wrote, is AI companies actually delivering the benefits they've promised — he specifically mentioned curing cancer — rather than arguing about messaging. That's a notably blunt admission for a company reportedly on pace for $65 billion in annual revenue and reportedly prepping an IPO as soon as this fall.