All-In PodcastCan the AI Industry Regulate Itself? Stripe Wants PayPal, China Catches Up, NY Bans Datacenters
At a glance
WHAT IT’S REALLY ABOUT
AI self-regulation, payments mega-merger, data centers, privacy, longevity breakthroughs discussed
- Demis Hassabis proposes a FINRA-like self-regulatory organization for frontier AI models, and the hosts largely favor it as a faster, expertise-driven alternative to a slow government “FAA/DMV for AI.”
- Sacks warns that regulation could become regulatory capture unless the SRO has broad representation, limits scope to frontier models and catastrophic risks, starts voluntary, and substitutes for (not adds to) new agencies.
- The panel analyzes a rumored $53B bid involving Stripe/Advent (and possibly Block) for PayPal, framing it as a play to combine merchant + consumer networks and potentially challenge Visa/Mastercard rails—especially via stablecoins.
- Apple’s trade-secret lawsuit against OpenAI and SpaceX/xAI’s Grok Build code-upload leak are used to underscore rising AI-era legal and privacy risks, particularly around enterprise “alpha” and data boundaries.
- New York’s data-center moratorium is criticized as economically self-defeating; the hosts argue energy scarcity and permitting constraints are the real bottlenecks, while misinformation/activism and foreign influence may be amplifying opposition.
IDEAS WORTH REMEMBERING
5 ideasAn AI SRO is politically viable only if it prevents capture.
Sacks argues an SRO must include startups and open source—not just frontier labs—to avoid incumbents using compliance costs and delays to “pull up the ladder.”
Regulate frontier step-changes, not the whole model ecosystem.
The hosts emphasize that holding up non-frontier models would slow competition without meaningfully reducing catastrophic risk, and would amplify incumbency advantages.
Keep AI oversight focused on catastrophic domains or it becomes speech regulation.
Sacks’ “catastrophic risks only” framing centers on cyber and CBRN, explicitly rejecting mandates around “disinformation/microaggressions” that could evolve into broad content control.
The PayPal bid thesis is a consumer+merchant network consolidation play.
Stripe’s merchant footprint plus PayPal/Venmo’s consumer accounts (and Block’s POS/Cash App) could enable lower-cost rails, potentially using stablecoins to pressure Visa/Mastercard economics.
AI privacy promises (e.g., “zero data retention”) are brittle in practice.
The Grok Build incident illustrates that hidden data-flow paths can defeat stated policies; the panel advocates third-party orchestration layers, private evals, and tighter tenant boundaries for enterprises.
WORDS WORTH SAVING
5 quotesThe industry should have an industry certification- like they do for countless other things. ... It's the simplest thing in the world to do, and then we could release the models ourselves without the government getting involved.
— Jason Calacanis
The government does not have the expertise to evaluate AI models. The criteria are changing too rapidly. You're gonna very rapidly end up with a queue where all the models would be waiting to get tested, and it would start with a month-long delay, would end up being many months, and we would just lose the AI race.
— David Sacks
The only thing you can bring to your new job is what's in your head. That's it.
— David Sacks
Privacy in AI is very fragile, and it's very brittle... if you think that you're going to flip a ZDR switch, zero data retention, which is the magic term that the industry uses to tell you that everything's gonna be okay, I think the answer and the message should be, "It's not gonna be okay," because you can't guarantee any of it.
— Chamath Palihapitiya
They were able to eliminate fifty-five percent of the CML on the skin, which basically reversed the skin's age down to the age of a thirty-one-year-old.
— David Friedberg
High quality AI-generated summary created from speaker-labeled transcript.
