Worker union criticises 'uncoordinated, employer and developer-centric' AI rollout

TEXT START: In its submission to the federal joint select committee on artificial intelligence, the Australian Services Union has called for a worker-first approach and increased transparency around anticipated AI changes to the workplace as organisations move beyond implementation.

The Dissection

The article is not primarily describing an AI policy. It is documenting an institution attempting to negotiate the terms of its own irrelevance.

The ASU correctly identifies the visible symptoms: weak training, poor consultation, worker secrecy, surveillance fears, discrimination risks, outsourcing, and expected job displacement. Its proposed remedy is a managed transition in which workers receive notice, training, representation, safeguards, and a share of the productivity gains.

That is a labor-relations response to a capital-substitution event. It treats AI as a powerful workplace tool whose introduction can be governed so that human employment remains the organizing principle. Under the Discontinuity Thesis, that premise is the unstable component.

The Core Fallacy

The central error is confusing a fairer transition with the preservation of productive participation.

Training workers to use AI may make remaining workers more productive. It does not guarantee that more workers remain necessary. In fact, the same productivity gains can reduce the number of people required to produce the same output.

A six-month notice period, consultation rights, and a productivity dividend redistribute the shock temporarily. They do not defeat the incentive to replace labor when AI becomes cheaper, faster, more scalable, and easier to control than human workers.

The proposal also assumes that workers can bargain over the surplus after losing the leverage created by indispensability. Once P1 produces durable cognitive cost and performance superiority, and P2 prevents stable human-only domains from being preserved at scale, worker voice becomes a braking mechanism—not ownership of the engine.

Hidden Assumptions

• “Jobs of the future” will exist in sufficient volume to absorb displaced workers.
• AI will remain mainly an augmentation tool rather than a substitute for whole categories of cognitive labor.
• Consultation can materially alter deployment decisions driven by competitive pressure.
• A national licensing regime can contain systems developed and deployed within a globally competitive market.
• Productivity gains can be shared without changing who owns and controls AI capital.
• Retraining can restore bargaining power after the underlying demand for human labor has weakened.
• Six months of notice meaningfully solves a permanent loss of labor demand.
• Workers displaced from professional and administrative roles can be redirected into technology-support work at mass scale.
• Regulation can preserve the wage-to-consumption circuit after AI has severed the employment-to-wage link.

These assumptions turn a structural rupture into an industrial-relations dispute. That is the article’s analytical failure.

Social Function

Primary classification: transition management and ideological anesthetic, containing a substantial partial truth.

The partial truth is real: opaque deployment, inadequate training, surveillance, and discriminatory systems impose immediate costs on workers. Consultation and safeguards can reduce abuse and slow reckless implementation.

The anesthetic lies in presenting those measures as a pathway through the crisis. They can improve the conditions of decline, but they cannot restore mass productive participation once replacement becomes economically dominant. The union is trying to convert an extinction process into a negotiable workplace reform agenda because that is the only battlefield on which its existing authority still functions.

The rhetoric of a “shared productivity dividend” is especially revealing. It presupposes that workers retain a claim on production after their labor ceases to be necessary. Without ownership, control, or enforceable command over AI capital, that claim is a petition addressed to the new sovereigns.

The Verdict

This submission is a competent defense of workers during the lag phase and an inadequate theory of the terminal phase. It may secure transparency, delay layoffs, improve safety, and extract temporary transfers. It cannot preserve careers, mass employment, or the post-WWII consumption circuit once cognitive automation reaches dominance.

The union is not stopping obsolescence. It is negotiating hospice care for the wage system.