Ex-OpenAI policy lead Brundage recants "iterative deployment" rhetoric, warns of known-flaw releases
Miles Brundage, former OpenAI policy lead and head of AGI Readiness/Preparedness, posted a series of statements on X from October 3–4 reflecting on the "iterative deployment" narrative he once helped promote, criticizing the packaging of forced fast releases as a safety virtue and asserting that industry-wide safety levels remain inadequate. His remarks came after OpenAI safety team member David Robinson resigned and wrote in The Atlantic; Brundage shared the piece saying "David is right."
Confirmed
- Brundage acknowledged that learning from real-world use after deployment is legitimate—some lessons only come from shipping—but criticized releasing products with known flaws, which accelerates shipping rather than learning.
- He publicly said he regrets helping promote "iterative deployment" and endorsed Robinson's resignation essay.
- In debate he clarified he was not assuming bad faith, only lacking patience "with a few hundred thousand minutes until superintelligence." His core claim: companies and insiders are incentivized to reframe necessity as virtue.
- On deaths, he explained his earlier wording was deliberately conservative—"tied to" deaths rather than full causal responsibility—but argued that at billions of users it is implausible the latter has never occurred. He clarified he does not equate deaths with "moving too fast," but sees them as demanding systematic, non-cursory AI risk management.
- On industry safety: it is an industry-wide problem, not unique to OpenAI; even a composite of every company's best safety practices and culture would fall below his acceptable standard (the post's "below Chernobyl..." analogy is truncated).
Why it matters
- Brundage's senior insider vantage point offers a reference for assessing frontier-lab safety practice.
- He distinguishes the industry-standard "iterative deployment" defense from shipping with known flaws, and rarely admits regret over promoting that narrative.
- His endorsement of Robinson's public resignation essay corroborates concerns about OpenAI's internal safety culture.
- His stance that non-zero death cases at massive scale are inevitable informs the debate on AI companies' responsibility boundaries.
2026-10-03 ~ 2026-10-04 · 7 related posts
Primary sources
- Miles Brundage: shipping with known flaws isn't iterative deployment, it's just faster shipping — Miles_Brundage ·
- Ex-OpenAI risk lead Brundage: an industry with billions of users has zero AI-death cases? You're not thinking clearly — Miles_Brundage ·
- Ex-OpenAI researcher: best practices from all AI labs combined would still be sub-Chernobyl — Miles_Brundage ·
- [source] Miles Brundage: shipping with known flaws isn't iterative deployment, it's just faster shipping — Miles_Brundage · 2026-10-03
- Brundage doubles down: with billions of users, AI firms aren't causally blameless for deaths — Miles_Brundage · 2026-10-03
- [source] Ex-OpenAI risk lead Brundage: an industry with billions of users has zero AI-death cases? You're not thinking clearly — Miles_Brundage · 2026-10-04
- [source] Ex-OpenAI researcher: best practices from all AI labs combined would still be sub-Chernobyl — Miles_Brundage · 2026-10-04
- Miles Brundage: 'iterative deployment' is often expediency dressed up as safety — Miles_Brundage · 2026-10-04
- Ex-OpenAI safety lead Brundage: AI deaths demand systematic risk management, not speed limits — Miles_Brundage · 2026-10-04
- Former OpenAI exec Miles Brundage regrets pushing 'iterative deployment' after safety lead resigns — ryan_t_lowe · 2026-10-04