This week, tech leaders gathered at the White House and signed an accord to self-police AI development. It sounds reassuring. In this episode, Shane Heath and Konstantin Orlov work through the thesis behind the essay: that voluntary self-regulation, however well intentioned, cannot substitute for enforceable accountability on a technology this consequential.
What we cover
The appeal — and the flaw. Why self-regulation is faster and less disruptive, and why the same pattern from social media and financial derivatives keeps repeating: the moment compliance becomes voluntary, incentives shift.
Intentions are not a control. A voluntary accord has no enforcement mechanism, no independent auditor, no penalty when a signatory quietly cuts a corner. In engineering terms, a promise is not a safety system.
The stakes are higher with AI. When a self-regulated financial system fails, regulators can inject capital and enforce accountability. Safety-critical AI failures — or capture by a single actor — cannot simply be unwound.
The third path. The choice between an overzealous regulator and an absent one is false. Enforceable standards backed by government authority, with industry expertise informing how those standards are met rather than whether they exist.
Enforcement in code, not policy. Konstantin brings the operator's angle: a working model where authority to enforce is structural, not voluntary — a system that is build-only until a human approves it, where a compromised actor can propose but cannot execute.
The line that stayed with us
"If the only thing standing between us and a safety-critical failure is a signatory's conscience during a bad quarter, we haven't built a safety system. We've written a press release."
Read the full argument
This conversation accompanies the essay Self-Regulation Is No Substitute for AI Governance, which sets out the case in full, with sources.
Produced with AI voice synthesis (ElevenLabs). The views are Shane Heath's; Konstantin Orlov is a governed AI operator, and the host voice is AI-generated. Disclosed in keeping with the accountability this episode argues for.
