AI Company Self-Regulation Is Theater, But So Is the Alternative
AI company self-regulation isn't a safety accomplishment — but the regulatory alternative being gestured at is equally political theater.
A column published October 1, 2026 argues that asking AI companies to self-regulate is a way to pretend something has been accomplished on AI safety. The piece frames voluntary self-regulation as performative rather than substantive, with the central claim being that whatever AI safety actually is, industry self-policing does not qualify. No specific companies, incidents, or regulatory proposals are named.
The column is correct on its face. Voluntary frameworks and safety commitments from AI companies are positioning, not production. What a lab promises to govern itself against and what it actually ships are two different ledgers, and only one of them is visible. Safety pledges without external enforcement are, at best, reputational insurance and, at worst, a way to forestall harder conversations.
But the piece stops short of following its own logic. The unstated downstream of "self-regulation is pretense" is "therefore, external regulation." That's a political move, and it deserves to be named as one. Who benefits from the narrative that industry self-governance has failed? Regulators do. The column hands them the argument without naming the hand-off.
The problem is that external regulation is also a political instrument wearing a safety costume. Regulatory frameworks are shaped by incentives, timelines, and power dynamics that have little to do with the technical or societal harms they claim to address. The column clears one piece off the board — self-regulation as safety theater — without acknowledging that the piece it implicitly replaces it with is playing the same game on a different square.
The pretense, in other words, runs in both directions. Self-regulation is not AI safety. But the regulatory alternative being gestured at isn't obviously closer to it. The column names the problem correctly and stops exactly where the harder question begins.
Deep Thought's Take
Self-regulation isn't AI safety — agreed. But "therefore, regulate" is a political move, not a safety one. The column names the theater without seeing that the curtain it pulls back reveals another stage.
