Author: Ben Thompson Publication: Stratechery Date: November 1, 2023 URL: https://stratechery.com/2023/attenuating-innovation-ai/
"Attenuating Innovation (AI)" is a November 1, 2023 Stratechery essay by Ben Thompson arguing that Biden's Executive Order 14110 on artificial intelligence functions as regulatory capture. Thompson contends that the order's compute-threshold reporting requirements apply only to the largest existing labs and so entrench incumbents while suppressing downstream innovation. The essay is a reference document for the regulation-as-incumbent-protection position.
Summary of argument
Thompson's central claim is that genuine innovation requires embracing uncertainty about the future, recognizing that people invent use cases continuously, and editing after invention rather than before. He argues that EO 14110 does the opposite by proscribing a future no one can yet see, which in his framing attenuates innovation rather than accelerating it.
The essay opens with the Bill Gates / Steve Jobs contrast from the 2007 D5 conference: Gates predicted the future confidently and, in Thompson's account, missed it, while Jobs said "I don't know" and shipped the iPhone. Thompson reads Microsoft's 2019 narrative blaming the antitrust case for the company's loss in mobile as a consumer-tech analog of regulatory-capture mythmaking, in which successful incumbents blame regulation for their own paradigm failures and then lobby for regulation that locks in their position.
Key claims
Thompson offers what he calls a "cynical read" of the AI-alarm wave: the May 2023 Center for AI Safety letter, signed by Sam Altman, Geoffrey Hinton, Yoshua Bengio, and others and comparing AI risk to "nuclear weapons and pandemics", supplied Washington the political cover to act (Center for AI Safety). The result, in his account, was an executive order whose compute-threshold reporting requirements (10²⁶ FLOP) apply only to the largest labs, that is, the existing oligopoly.
He frames the incumbent-protection mechanism as three steps: define the regulated category narrowly enough that only existing leaders qualify; layer on disclosure, safety-testing, and red-teaming requirements that carry meaningful compliance overhead; and let that compliance overhead become a moat that startups cannot afford to cross to enter the regulated tier. In Thompson's reading, the framing of AI as nuclear-weapon-class supplied the political will and the compute threshold supplied the moat.
Thompson argues that open-source models are both safer, because they avoid concentrating capability in a few stewards, and more conducive to innovation, because they place no regulatory chokepoint at the model layer. The essay serves as a foundational anchor for Open-Source AI / Open-Weight Models.
Reception and related positions
In Anthropic and Alignment (March 2026), Thompson explicitly references this essay as the foundation of his anti-Anthropic-position argument.
The essay's position runs counter to The AI Grand Bargain, in which Buchanan and Collins argue the laissez-faire model is exhausted; Thompson, by contrast, argues the laissez-faire model has barely been tried because the EO already locked in the wrong path. How MAGA Learned to Love AI Safety describes the populist-conservative coalition that defeated the state-regulation moratorium, which echoes Thompson's anti-regulatory-capture frame from the opposite end of the political spectrum.
Provenance
Published on Stratechery on November 1, 2023. The piece is an opinion essay; its claims about regulatory capture and incumbent protection are arguments advanced by Thompson rather than established findings.
Relationships
- depends-on: Ben Thompson
- related: Executive Order 14110 (RESCINDED) — primary subject of critique (Biden Oct 2023 AI EO)
- related: Sam Altman — CAIS-letter signatory
- related: Geoffrey Hinton — CAIS-letter signatory
- related: Yoshua Bengio — CAIS-letter signatory
- related: Open-Source AI / Open-Weight Models — foundational anchor for this concept page
- related: Center for AI Safety (CAIS)
- contradicts: The AI Grand Bargain — Ben Buchanan and Tantum Collins (Foreign Affairs, October 2025)
- supports: Anthropic and Alignment — Ben Thompson (Stratechery, March 2026)