EU workers' rights missing from landmark AI law

MEPs inched further on labour rights than the original European Commission proposal.

But in the race to legislate potentially disruptive technologies, EU institutions are still largely overlooking the workers affected by incoming AI systems, unionists say.

AI is changing workers' lives in many ways.

It's already used for tracking drivers in apps such as Uber and Deliveroo, in helping robots work alongside people on factory lines, and in managing patients' medical data for doctors and nurses.

The EU Parliament has added a potential ban on risky AI emotional-recognition systems, which can warp the employer-employee relationship — but that's as far as progress on employment issues goes.

"We need a specific directive on labour relations because there are many question marks that are not clarified in the AI law because that was not its [original] purpose," said Aída Ponce, a senior researcher at the European Trade Union Institute (ETUI) in Brussels.

"Specific provisions are needed to set a clear and concrete limit to the use of AI systems in industrial relations, up to the point of monitoring," she said.

The ETUI's sister organisation, the European Trade Union Confederation (ETUC) in Brussels, has also said a separate AI-workers directive will be needed.

And it specifically criticised as a "major loophole" that only AI applications categorised as high-risk are regulated.

According to article six of the EU Parliament's position, they can only be restricted in the workplace if they pose a "significant" risk.

But while "[AI] providers will generally state that their applications may pose risks to workers, [they] will classify them as non-significant," ETUC warned.

"This additional burden on workers is unacceptable and leaves their safety and rights open to abuse," it said.

In the EU Commission's original AI Act proposal, providers will have to assess the potential risk of their own technologies before launching them on the market.

Depending on the level of risk, EU regulators would then impose restrictions, including up to bans for the most dangerous products, such as AI social-scoring — classifying people according to personal characteristics and social interactions.

For IndustriALL, a global union representing 50 million workers, AI systems should be classified as high-risk whenever they use workers' personal data.

IndustriALL also says decisions made by AI must be legally "traceable" and "reversible".

AI facts on the ground

But amid the unions' alarm, market developments are, in any case, galloping ahead of the EU legislative process.

AI systems are already replacing workers — even in the EU institutions.

Over 40 percent of administrative and legal services could soon be automated, according to US-based investment bank Goldman Sachs.

The use of AI in the workplace increased by almost 30 percent between 2018 and 2023, according to a survey by US management-consultancy firm Boston Consulting Group.

Almost a third of respondents said they were worried about being replaced.

The EU Commission's main response on workers' risks so far has been: more skills, with EU press material fond of buzzwords such as "upskilling", "reskilling", and "digital skills".

But for the ETUI's Ponce, that's missing the point on the potential damage that bad AI can do to people's working lives.

"The focus on the skills policy needed to counter the use of AI is misleading," she said.

"The approach needs to be completely reframed," Ponce said.