Sacks vs. Anthropic: Safety, Moats and Conflicts

The fight between David Sacks and Anthropic is the best live example of a real argument made from the wrong chair. The argument itself is legitimate. AI-safety discourse can harden into a moat, where leading labs push for rules that slow everyone except themselves. Someone should test whether a specific doom scenario is load-bearing or just load-collecting, and Sacks is willing to pay the social cost of asking.

The problem is his position. As I described on the show, he's been pressing this case from inside government, with influence over things like export controls and procurement, while his venture firm holds stakes in companies that benefit when a competitor is constrained. Anthropic disputes his framing of specific incidents, and I treat those as contested. But the structural point doesn't depend on any single incident. A conflict of interest doesn't require bad faith. It's a reason for more scrutiny, not less. That's why I called it regulatory arbitrage with a podcast.

The irony I noted on air is that Founders Fund, co-founded by Sacks's longtime collaborator Peter Thiel, has invested heavily in Anthropic. The capture story gets messy when the capital moves through the same circles.

I also think Sacks sincerely believes the softer version: not that safety is a psyop, but that the field has become a self-serving industry. The trouble is that once you believe the output is rigged, you stop distinguishing "the research is wrong" from "the research is fake."

My lean on his core move is still no. But it's a lean, based on a first-pass scorecard that read only one All-In transcript in full. If he applies the same scrutiny to Craft-friendly companies, or publicly corrects a claim he made about Anthropic, I'll update here.