For most of the past decade, the dominant assumption in AI-and-labor debates has been that technological capability would determine outcomes. If a task could be automated, it would be — on the employer’s timeline, at the employer’s discretion, with workers and regulators left to respond after the fact. Institutions have largely failed to keep pace. But a different kind of response is starting to take shape — not in legislatures or regulatory agencies, but at the bargaining table.
A January 2026 report from Connected by Data documenting their work with the UK’s Communications Workers Union (CWU) describes something structurally significant: a union developing a formal model collective agreement designed to govern AI deployment across its employers. At its core is a proposed Joint Union-Management AI Committee (JAIC) — a body with equal worker-employer representation and binding authority. Under the proposed terms, no high-risk AI system could be procured or deployed without the union’s explicit agreement.
The model agreement is built around three explicit negotiating anchors. First, worker voice at all levels of tech decision-making — from design through deployment, not just after the fact. Second, a fair share of AI-related benefits: pay, skills development, working hours, and training. Third, robust safeguards against bias, discrimination, and other algorithmic harms embedded in the contract itself, not delegated to management discretion.
The CWU framework didn’t emerge in a vacuum. The Writers Guild of America’s 2023 strike produced the first enforceable workplace agreement on generative AI, establishing that collective bargaining could set binding limits on how studios deploy the technology. A July 2025 ILO working paper explicitly frames collective bargaining as a primary mechanism for AI governance. Littler Mendelson, a major US labor and employment law firm, has advised unionized employers to prepare to bargain over AI adoption as a standard labor relations obligation. And a peer-reviewed article published in Information, Communication & Society in June 2025 develops the concept of “worker-led AI governance” to describe this emerging pattern.
What makes the CWU model notable is the template logic. A replicable agreement, stress-tested with academics and lawyers and designed to travel across sectors, is infrastructure — not just a one-off negotiation win.
Technological capability sets the ceiling for what automation can do. Bargaining structures increasingly determine what it will do.
Sources: Connected by Data / CWU, January 2026 — Full report | ILO Working Paper 144, July 2025 — Full report | Information, Communication & Society, June 2025 — Full article
Deja un comentario