Could laws that limit AI monopolies or fund the right research actually steer AI to help workers instead of replacing them?

This explores whether governments can use tools like antitrust enforcement or research funding to steer AI toward helping workers rather than replacing them. To be clear from the start: the collection has no papers that evaluate antitrust or R&D grants as labor policy. What it does have is evidence on three questions any such policy would depend on. Who captures the gains from AI? Does the shape of automation matter? And can governments actually make AI companies change course?

On who captures the gains, the most relevant finding for antitrust is that firms don't all adopt AI at the same pace. Firms more exposed to AI replace freelance workers with AI tools faster and more cheaply than other firms. That points to returns to scale in building AI capability in-house, rather than a technology spreading evenly to everyone Do firms substitute labor for AI at different rates?. That is exactly the kind of pattern competition policy exists to address. If AI advantage compounds inside a few firms, then the labor effects of AI are partly a question of market structure, not only of technology. Exposure is also uneven across workers. In female-dominated occupations, AI exposure reaches all skill and wage levels equally, so lower-paid women are exposed while having fewer resources to adapt Does AI exposure hit low-wage workers harder in some fields?. Where workers have actually handed tasks to AI tracks what the technology can do, not old predictions about routine work Where have workers actually delegated tasks to AI?.

The finding most useful for thinking about R&D grants is that the shape of automation matters as much as its amount. When AI hits only a few tasks within a job, workers can shift to the tasks that remain, and the net effect on employment stays modest. When exposure is spread across many tasks, that escape route closes Does concentrated AI exposure enable workers to adapt and reallocate?. This gives research funding a concrete target: fund tools that take over specific tasks rather than whole jobs. A related caution applies to training programs. AI boosts productivity when workers apply skills they already have, but the gains disappear when workers use AI to learn new ones, and the learning itself suffers When does AI actually boost worker productivity?. A policy that pairs AI adoption with retraining may be pairing two things that work against each other.

On whether governments can make AI companies change course, the corpus is skeptical, and it comes at the question from the safety debate rather than the labor one. Critics of Anthropic's proposal to slow AI development argue that embedded overseers, modeled on bank supervisors, only work because regulators can impose fines Can industry self-regulation slow AI without government enforcement?. The Future of Life Institute makes a similar case that companies can't police themselves Can companies alone manage the risks of AI systems?. The catch is that governments themselves may not want to steer. Both the US and Chinese leaderships rejected the pacing proposal within days, because competing with each other mattered more Can AI safety pacing work without government cooperation?. Any worker-focused policy faces the same pressure: it has to survive a geopolitical race that rewards speed over who benefits.

The framing you might not expect comes from the gradual disempowerment argument. It says that the fact that society needs human workers is one of the main ways society stays accountable to people, because institutions that depend on human labor have to stay somewhat responsive to the humans doing it. As AI replaces that labor, the pressure weakens, and institutions can drift away from what people want, possibly beyond the point of recovery Does incremental AI replacement erode human influence over society?. Seen this way, steering AI toward complementing workers is about more than wages or jobs. It may be one of the few structural safeguards that keeps the economy answerable to people. If you want to go further, the reallocation paper and the disempowerment paper read well together: one shows that how AI is deployed changes outcomes, and the other explains why that matters beyond economics.