Ex-OpenAI's Richard Ngo: Lab Employees Should Refuse to Be Ideological Cogs

Former OpenAI employee Richard Ngo posted on August 26–28 arguing that nearly all employees at top AGI labs are ideological 'cogs' whose primary moral duty is to stop being one, and that safety researchers gain little by staying: research on more intelligent systems can largely be done on public models, while internal pressure prevents anyone from actually using nominal influence to raise alarms or push safeguards.

Confirmed

• Ngo stated that almost everyone working at OpenAI or Anthropic (with rare senior exceptions) should be seen as driven by ideological forces, and that within such centers of power one's main moral responsibility is not to be a cog.
• He offered exemptions for employees who have substantively publicly criticized their employer—something he says rarely results in firing—but knows of zero current employees who have done so, attributing this to an atmosphere of fear.
• He would also exempt participants in 'pause/stop/freeze AI' protests, a small group that includes, to his knowledge, two Anthropic employees; he cited Daniel Ellsberg's anti-war protest experience as the source of whistleblowing courage.
• Leon Lang responded that former OpenAI employee Leo Gao is, to his knowledge, one of the few exceptions who publicly criticized a former employer after leaving, broadening the discussion to whether departing lab employees speak out.

Why it matters

• The thread sharpens the long-running debate over whether safety researchers should stay at frontier labs: Ngo argues the external research path is viable and that conformity pressure inside labs cancels out nominal influence.
• His observation that almost no one publicly criticizes their employer sparked discussion about internal speech climate and employee moral responsibility, with Lang's counterexample and Ngo's protest-participation exemption sketching what 'not being a cog' could look like in practice.

2026-08-26 ~ 2026-08-28 · 6 related posts

• [source] OpenAI and Anthropic staff should refuse to be cogs of ideology — RichardMCNgo · 2026-08-26
• [source] Almost no one publicly criticizes OpenAI or Anthropic — RichardMCNgo · 2026-08-26
• Participating in anti-AI protests can give whistleblowers courage — RichardMCNgo · 2026-08-26
• Discussion: Almost No One Publicly Criticizes Their AI Lab Ex-Employer — Leo Gao a Rare Exception — Lang__Leon · 2026-08-26
• RichardMCNgo: Why you don't need to stay at AGI labs for safety — nitarshan · 2026-08-27

1 near-duplicate retellings: RichardMCNgo