The US should not regulate most AI, letting the tech mature for positive uses, to avoid favoring incumbents, killing startups, and forcing the industry overseas
that we've been living in an era of overregulation, which harms innovation and ossifies incumbents — is spot on But I disagree with his first example, regarding the Telecommunications Act of 1996 (TCA) Bill Gurley / @bgurley : Thanks to @friedberg & @theallinpod for inviting me to speak. One key phrase from Stigler is so important, “... leading to a net loss for society.” It's not just that these lobby efforts have undo influence, its that this influence causes specific harm to society. @apompliano : “The reason Silicon Valley has been so successful is because it is so f**king far away from Washington DC” [video] Sid Powell / @syrupsid : Worth watching @bgurley. Whenever i attend a crypto conference, I hear that institutions are “waiting for regulation”. Why? Well, watch this Bill Gurley's insights on regulation resonate deeply. At crypto events, we're told “more regulation = more trust.” But new rules often... Misha Saul / @misha_saul : GPs now largely exist as a function of regulatory hangover + social support nodes (not health experts) “models such as Med-PaLM2 from Google now outperform medical experts to the point where training the model using physician experts may make the model worse.” [image] @theallinpod : All-In Summit: 2,851 Miles with @bgurley “The best talk in the history of All-In... and we need to get it out there immediately so it can start going viral.” - David Sacks, @allinsummit 📷 https://www.youtube.com/... LinkedIn: Allan Thygesen : Fantastic provocative talk by legendary VC Bill Gurley on regulatory capture in the US. Really worth spending 25 min to watch and listen. … David Roberts : If this speech by Bill Gurley gets you fired up, I encourage you to get in the (political) arena. Thanks to David Friedberg, Chamath Palihapitiya …
Context & Ripple Effects
This is an early articulation of the argument that AI rules can become a barrier to entry: compliance costs and lobbying capacity may advantage firms already at scale. It sits opposite calls for rapid public protection, including Yoshua Bengio’s call for governments to act quickly.
Later coverage shows the split persisted: some industry leaders urged the US not to rush into an EU-style approach, while critics of California’s proposed safety regime argued it could burden open-source, academic, and public-sector work.
First-order effects
- The intervention strengthens the anti-regulation case in the AI policy debate, centering startup formation and domestic development rather than only model-risk controls.
- It directly challenges AI companies and policymakers advocating broad rules to show that proposed obligations are proportionate and do not chiefly reward established compliance teams.
Second-order effects
- Debate over AI oversight becomes a contest over market structure: incumbent firms can portray rules as safety measures, while challengers can portray them as entry barriers—a conflict later captured in warnings that regulation could lock in early AI winners.
- Developers of open models, universities, and public-sector users gain a common argument against sweeping requirements, particularly where rules could attach heavy duties to smaller deployers.
Third-order effects
- If AI governance is designed primarily around frontier-firm capabilities, compliance may increasingly function as market access—concentrating influence among companies able to absorb legal, technical, and reporting costs.
- The enduring policy challenge is likely to be targeted safeguards rather than a binary choice between no rules and broad preemptive regulation; the corpus documents continuing disagreement over where that boundary belongs.
The trend: AI governance is becoming a market-structure debate in which safety regulation, startup competition, and national development are increasingly treated as inseparable.