Last week, journalist Serafí del Arco highlighted in his weekly column "Counteroffer" on eldiario.es a report that had gone largely unnoticed amid the media noise. It is titled "Industrial Policy for the Intelligence Age: Ideas to Keep People First," published by OpenAI, the company behind ChatGPT. In its thirteen pages, one of the most influential companies in the advancement of Artificial Intelligence (AI) proposes a new social pact: to redistribute the wealth generated by automation, tax automated capital, and establish 32-hour workweeks without salary reductions. In previous industrial revolutions, it took decades to create mechanisms that balanced the relationship between capital and labor: unions, collective bargaining, the right to strike, and other labor rights. The fact that a key company in this transformation recognizes the need for political intervention is significant. However, beyond the surprise, what is relevant is not only what the report admits but also what it leaves unresolved: who will be responsible for negotiating these changes? The document has undeniable value: it brings to the table issues that many prefer to suggest rather than assert. The market alone will not manage the technological transition. AI is not just another tool; it represents a historical scale change that, without public rules and strong institutions, will concentrate benefits in a few and shift costs to the majority. The report itself warns of the risk that wealth will concentrate in a small group of companies, which is paradoxical coming from an organization that was founded for non-profit purposes and is now a central player in technological capitalism. The underlying debate, however, is not about OpenAI's intentions but about the impact on the labor market. The report mentions "efficiency dividends": if automation allows for the same output in less human time, that freed time should translate into a better quality of life for workers, not just higher corporate margins. Fewer working hours, more rest, and better work-life balance: the distribution of progress. The proposal is sensible. If managed well, AI can reduce repetitive tasks, improve public services, enhance professional capabilities, and increase productivity. But the same technology can also be used to intensify workloads, increase control, reduce autonomy, and obscure the relationship between effort and reward. The ideal of abundance can degenerate into a new version of Taylorism, now managed by algorithms. This is not a hypothetical scenario: it is already happening in warehouses where every movement is digitally monitored, in customer service centers where every second is measured, or on platforms where algorithms decide hiring, schedules, and salaries. AI does not enter a neutral ground but rather into labor relations already marked by inequalities. This is where unions face a historic challenge. Collective bargaining emerged to correct the asymmetry between those who organize work and those who perform it. In previous industrial revolutions, it took decades to build tools to balance that relationship. Now, the speed of technological change is much greater, and its effects on employment and working conditions are already palpable. Therefore, the response cannot be limited to resistance. It is necessary to prevent layoffs disguised as innovation, avoid automation that precarizes work, and demand transparency in the algorithmic systems that evaluate and manage people. But this is not enough. The real challenge is to participate in designing the change, not just managing its consequences. There is a fundamental difference between negotiating the effects and participating in the decisions. The former mitigates damage; the latter distributes power. This should be at the center of upcoming collective agreements: rights to information and consultation, independent audits, traceability of automated decisions, effective human oversight, and the ability to veto violations of labor rights. There is also a structural dimension: if AI reduces direct employment or shifts wages toward capital, it also erodes the contributory base that sustains the social state: pensions, unemployment, public health. It is not enough to ask that "machines contribute"; it is necessary to rethink how social protection is financed when added value is concentrated in profits, intellectual property, and automated capital. AI is already present in workplaces. The question is not whether it will enter, but under what conditions it will do so and whether, when that happens, workers and their unions will have the strength and representation necessary to ensure a fair transition.