Ever since the AI hype began with the release of ChatGPT in November 2022, the EU has moved from wanting to set global regulatory standards with its AI Act to being worried about falling behind the US and China in the tech race. In the US-EU trade deal inked just a few days ago, for instance, the EU commits to purchasing €40 billion worth of AI chips ‘essential to maintaining the EU’s technological edge’. Its AI Continent Action Plan unveiled in April painted a compelling vision of Europe as a global leader in artificial intelligence, backed by €200 billion public-private investments, €20 billion for 5 gigafactories and an EU single market for data. But throwing money at the problem won’t solve the EU’s AI woes; the bloc’s strength lies in the European model’s ability to shape technological transformations for the public good and ‘aligned with EU values’, as the authors of the plan acknowledge. Strangely enough then, the AI plan fails to mention a central pillar of the European economic model: social dialogue between workers and employers to regulate AI the European way.
For decades, unions have negotiated protections in response to digitalisation — from early software systems to algorithmic management. AI is simply the next chapter in this ongoing story of workers and their unions shaping technology. Many of the seven million service workers that my organisation UNI Europa – a European trade union federation operating across 50 countries – represents are on the frontline of the AI transformation. Generative AI, in particular, threatens to displace jobs in administration, contact centres and the creative industries.
While the AI Continent Action Plan focuses on digital infrastructure – AI factories, gigafactories and data labs – it repeats a vocabulary of ‘trustworthy’, ‘human-centric’ and ‘democratic’ AI. But these adjectives are not self-executing goals; they must be anchored in workplace democracy, collective bargaining and social dialogue. A recent EU-wide poll demonstrated widespread support for the regulation of AI in the workplace. And European trade unions are demanding urgent action to regulate AI in the world of work — in the EU, the UK and beyond. But legislation alone cannot keep pace with AI’s rapid evolution. That’s where collective bargaining must step in.
Collective bargaining can be faster
AI doesn’t wait. Its systems are being deployed today in worker recruitment, scheduling, productivity monitoring — and even termination decisions. Meanwhile, Generative AI applications train on data generated by creative workers – from visual artists, performers, to filmmakers – often without compensation. Governments, despite their best efforts, simply cannot legislate fast enough to address these transformations in real time. The EU’s AI Act, which doesn’t even regulate the effects of AI in the workplace, is a case in point: it won’t be fully applied until 2027.
In contrast, collective bargaining is nimble.
For instance, years before any legislation arrives, it allowed remote workers exposed to AI to negotiate working time arrangements, helped content moderators mitigate psycho-social risks from watching gruesome videos, and ensured that biased data isn’t used to train a discriminating AI system. This way, collective bargaining ensures that AI deployment preserves a significant human presence at work — not just as passive overseers, but as autonomous and independent agents capable of supervising, contesting and overriding automated decisions. This is how our societies can operationalise human-centric AI almost in real time.
New AI applications will find more acceptance among workers if they are informed, consulted and negotiated with before any new technology is introduced.
AI systems and their implementation at work are often very technical matters. In many cases, employers buy solutions off the shelf without deeper technical understanding. As political economist Steven Rolf has argued, creating transparency, involving workers’ representatives in all stages of product purchase and implementation can help mitigate risks in an early stage. Workers and their representatives must be involved from the beginning of AI procurement and system design – not just after implementation – to ensure alignment with ethical principles and workplace needs. They are most closely exposed to AI systems; companies should use, not ignore, their expertise.
Likewise, new AI applications will find more acceptance among workers if they are informed, consulted and negotiated with before any new technology is introduced. This prevents rushed management decisions to buy hyped-up, but ultimately unnecessary technology that could end up being harmful to the company. Moreover, adequate evaluation mechanisms can help decide whether a new AI system really leads to more productivity and improves work processes for all.
In fact, this is already happening. Across sectors, workers are negotiating the impact of AI. Based on the example of their US colleagues, German actors have recently negotiated a collective agreement that sets new standards on transparency, consent and financial compensation for the use of generative AI. Deutsche Telekom’s works council co-authored an AI manifesto, while Italian customer service workers concluded a groundbreaking agreement regulating the experimental use of an AI tool named ‘Agent Assist’ in customer service operations. These examples show that collective bargaining is not just possible — it works.
We must stop unfair competition where companies undercut decent jobs by sourcing cheap data labour without rights or representation.
But collective bargaining on AI isn’t just about worker protection; it’s a competitive advantage. By involving workers in shaping how AI is deployed, companies gain legitimacy, avoid backlash and often arrive at better, more practical solutions. In a sector with high fluctuation and competition over talent, such as ICT, it helps retain high-skilled workers through offering quality jobs.
Moreover, while the promise of AI is productivity, efficiency and growth, these gains rarely trickle down to workers. We need a serious debate about redistribution: of time, income and power. Without it, AI risks reinforcing existing inequalities and fuelling discontent, rather than empowering workers. Collective bargaining is the process by which workers can have a say in the distribution of productivity gains generated by AI.
In short, collective bargaining ensures that decisions about workers are not made without workers. The European Commission rightly emphasises the need for speed, investment and innovation in its AI plan. It should also recognise that the same urgency must apply to social regulation — and that means promoting, enabling and, where necessary, requiring collective bargaining.
This includes using public funds and procurement rules to incentivise compliance with collective agreements, conditioning access to strategic investment support on social dialogue and integrating these requirements into any new initiatives. It also means supporting workers’ power in the ‘data value chains’ that fuel AI systems. This is especially true in outsourced and fragmented roles often located in the Global South or low-wage European countries like Poland, Romania, Portugal, Ireland and Spain. We must stop unfair competition where companies undercut decent jobs by sourcing cheap data labour without rights or representation.
Europe has always stood for a social model that combines innovation with inclusion, prosperity with protection. As we race to build AI supercomputers and deploy frontier models in a rapidly changing world order, let’s ensure that we don’t just make Europe an AI Continent, but make AI on our continent truly European.
