AI Company Employees Should Unionize

The case for safety-conscious employees at frontier AI companies organizing is that they have unusual leverage: a small number of highly skilled researchers would be difficult to replace, making collective action more practical than in industries with abundant replacement workers. Even an informal employee group could coordinate demands and, in some circumstances, strike without becoming a formally recognized majority union. Many safety concerns - such as whistleblower protections, staffing, audits, and the prospect that recursive self-improvement will eliminate researchers’ jobs - also connect to workplace conditions. Collective bargaining could help employees press for safer practices even when executives fear competitors will move first.

Unions could also protect researchers from questionable firings, demand follow-through on company safety commitments, and seek transparency about computing resources and AI progress. Separate unions at rival firms might coordinate around limits on risky research, reducing competitive pressure and potentially creating a basis for industry standards or government action. The argument acknowledges legal limits, especially for demands unrelated to workplace conditions, but says workers’ scarcity and the reputational cost of firing them provide leverage beyond formal legal protections. For employees who instead consider whistleblowing or quitting, the piece recommends consulting Psst, a nonprofit that offers confidential legal advice, warns that solo whistleblowing can expose workers to detection, and suggests that strategic negotiation or a public departure may be more effective. It urges action soon, on the grounds that AI could soon erode employees’ bargaining power.