The Workers Inside the Machine - Labor Rights & AI

For years, the technology industry sold its employees a special story about power.

You were not labor. You were talent.

You did not need a union because your skills made you mobile, your salary made you secure and your proximity to the future gave you influence. If management made a bad decision, you could leave. If a company crossed an ethical line, engineers could object from within. Expertise itself was supposed to function as bargaining power.

Artificial intelligence is exposing the limits of that arrangement.

Technology workers are organizing around layoffs, surveillance, workload intensification and the use of their work in military or politically controversial systems. Employees at companies including Google DeepMind and Meta are exploring collective action in an industry that has historically resisted it. The shift is still tentative, but the reason behind it is clear: AI is changing not only what technical workers build, but the terms under which they build it.

The same workers creating automation tools are watching their employers eliminate jobs, restructure teams and require fewer people to produce more output. Software engineers increasingly review machine-generated code rather than authoring every line themselves. Some are finding that the expertise that once made them indispensable is being converted into training data, workflow templates and supervisory labor.

This does not mean software engineering is disappearing. It means the balance of power inside the profession is changing.

A worker may still be highly skilled while becoming easier to monitor, compare and replace. AI can measure output, suggest staffing levels and standardize tasks that once depended on individual judgment. It can also allow management to claim that reductions are not strategic choices but unavoidable consequences of technological progress.

That claim deserves scrutiny.

Adecco argues that AI is not producing an economy-wide employment collapse and that some companies are using it to explain layoffs caused by ordinary restructuring or weak performance. Yet that does not make the disruption imaginary. It reveals that AI now performs two jobs for management: it can automate work, and it can legitimize decisions made for other reasons.

This is where collective bargaining becomes more than a dispute over wages.

Workers inside AI companies possess knowledge the public rarely sees. They understand how models are evaluated, where safeguards fail, which products are being rushed and how corporate incentives shape deployment. Individual whistleblowers can reveal pieces of that system, but they carry enormous personal risk. A workforce with negotiated protections could create a more durable form of internal accountability.

Collective bargaining could establish rights to disclosure before automation-driven restructuring, independent review of employee-surveillance systems, severance standards, limits on the use of worker activity as training data and channels for objecting to dangerous deployments without professional retaliation.

None of this guarantees enlightened governance. Unions can defend narrow interests. Technical workers may prioritize their own security over contract workers, data annotators and communities affected by AI infrastructure. A unionized engineer is not automatically a public representative.

But the alternative is to leave most consequential decisions to executives, investors and government officials while treating the people closest to the systems as replaceable implementers.

That is becoming untenable.

The AI governance debate usually focuses on regulations imposed from outside the company. Those rules are essential. But governance can also emerge from inside the production process, through people with the knowledge and institutional protection to say no.

The workers building AI do not control it merely because they understand the code.

They control it only when they possess the power to influence what happens next.

AI may eventually transform labor beyond recognition. Before it does, it may produce a more immediate reversal: the industry that promised to eliminate collective labor could help revive it inside its own walls.

The most important safety layer may not be another model watching the model.

It may be organized humans watching the company.