Australian unions demand AI guardrails in public sector contracts

Australian federal employee unions are pressing for binding AI safeguards in upcoming enterprise agreements that will govern working conditions for more than 100,000 public servants through early 2030, setting up a potential clash with a government eager to accelerate technology adoption.

The Public Service Commission this week tabled an initial offer that would establish rights to AI ethics training and require management to consider technology impacts on employment decisions. But the proposal stops short of demands from the Community and Public Service Union (CPSU), which represents 48,000 federal workers and wants explicit prohibitions on AI in performance reviews, hiring decisions, and other employment matters affecting staff.

Why it matters

The negotiations will lock in AI governance terms for Australia's federal workforce during a critical three-year window when the government plans to rapidly scale technology use across agencies. The outcome could establish precedents for AI workplace protections beyond the public sector, and the tension between productivity goals and worker safeguards mirrors debates unfolding in organizations worldwide.

Unions seek enforceable protections

CPSU National Secretary Rebecca Fawcett said the government's opening position falls short of its stated commitment to be a "model employer" on AI. Union members want guarantees that humans retain final decision-making authority on AI-assisted outcomes and that the technology won't be deployed in processes directly affecting employees without their consent.

Professionals Australia, representing engineers and scientists in government, is seeking mandated human oversight of AI systems plus legal protections for specialists who raise concerns about technology deployment. The Australian Services Union's Tax Office branch wants a firm rule that "all decisions that directly affect people must be made by a person and not AI or a bot."

The Public Service Commission countered that existing workplace provisions already address consultation and job security concerns, and warned that "prescriptive or overly detailed provisions" risk becoming outdated as AI capabilities and regulations evolve.

Productivity push drives government stance

Treasurer Jim Chalmers has identified AI adoption as a key lever for boosting productivity across Australia's economy. Productivity Commission Chair Danielle Wood noted that government services account for roughly 20 percent of economic activity when state and territory operations are included, making public sector efficiency gains significant for national productivity targets.

Finance and Public Service Minister Katy Gallagher's office said AI adoption aims "to lift productivity and improve services, rather than replace public servants," positioning the government as committed to workforce protection while pursuing technology gains.

The government has issued directives requiring agency leaders to consult employees where AI will have significant impact, and published a 2025 AI plan mandating safe deployment environments. But unions argue these guidelines lack the enforceability of enterprise agreement clauses, particularly given a federal election is due by May 2028 that could bring policy changes.

Limited leverage until 2027

Unlike previous bargaining rounds, current enterprise agreements haven't expired, restricting unions from organizing protected strikes until after February 2027 in most cases. Any escalation would require workers to delay their own pay increases to take industrial action.

Negotiations remain in early stages, and the final terms will determine AI governance for Australia's federal workforce through a period of rapid technological change.

These details were first reported by Miriam Webber for POLITICO.