Unions and Collective Bargaining over AI

Unions rarely bargain over 'AI' as such. They bargain over familiar subjects that AI touches: job security, workload, discipline, surveillance, training and pay. Because the subjects are familiar, AI clauses can plug into existing enforcement. A worker disciplined on the basis of an algorithm's score can file a grievance, and an arbitrator can decide whether the contract was violated. A few patterns recur. Notice-and-bargain clauses require the employer to tell the union before introducing a new technology and to negotiate over its effects. Job-protection clauses restrict layoffs caused by automation or require reassignment and retraining. Human-in-the-loop clauses say that decisions such as discipline, firing or clinical judgment must be made by a person. Surveillance limits restrict using monitoring data to discipline workers. Content clauses, common in media and entertainment, control credit, consent and pay when AI output or likenesses are used. Real examples span sectors. The 2023 WGA and SAG-AFTRA contracts set rules on AI-written material and digital replicas. The International Longshoremen's Association struck East and Gulf Coast ports in October 2024. The deal that followed limited full automation at terminals. Culinary Union contracts in Las Vegas in 2023 included notice and training terms for new technology. In 2023 Microsoft and the AFL-CIO announced a partnership to share information about AI with workers and include their views in its development. One misconception is that unions simply oppose technology. Many contracts accept new tools in exchange for protections and a share of the gains. Another is that these deals protect most workers. In the United States, only about one in ten wage and salary workers belongs to a union, so most workers have no contract covering AI. In countries with works councils or sector-wide bargaining, coverage is broader.