The surveillance paradox at the heart of France's AI push
On 15 May 2026, workers at Orange — France's largest telecom operator and a state-backed champion — walked off the job. Their grievance was not wages or redundancy terms. It was a system the company had built: AI software that tracks employee movements throughout the working day and algorithmically evaluates the content of their conversations. The workers called it what it is: digital surveillance.
The same week, the French government announced a €1.5 billion investment in quantum computing and advanced microchips. President Macron's administration framed it as a matter of strategic sovereignty — France building the AI infrastructure that will define economic competitiveness and military capability through the next decade. The announcements landed in business pages and tech feeds, drawing congratulations from the Élysée's communications team.
The Orange strike drew less coverage. But the two stories belong together, because they expose a contradiction at the centre of France's AI posture — and, by extension, at the centre of most Western industrial policy on artificial intelligence.
The state is wagering public money on AI as a strategic capability. The same state is allowing corporate actors to deploy AI as a mechanism of workplace control, without apparent concern for where the technology's boundaries should sit when it is pointed at the people who make the productivity possible.
France's sovereign AI project, and its domestic blind spot
The €1.5 billion commitment — spread across quantum research facilities, semiconductor design centres, and public-private partnerships with firms including Atos and Thales — is not a modest bet. It reflects a calculation, shared across Brussels and Washington, that AI infrastructure is too strategically sensitive to leave to overseas supply chains. France's national security apparatus sees quantum computing as a long-horizon necessity for cryptography, logistics optimisation, and signals intelligence. The microchip investment follows the pattern set by the EU Chips Act: reduce dependence on Taiwanese and South Korean foundries, build domestic fabrication capacity, keep the IP inside the bloc.
These are coherent objectives, and the industrial logic behind them is sound. What the policy documentation does not address — in any of the statements that accompanied the May 2026 announcement — is the domestic governance of AI systems inside French companies. The surveillance infrastructure at Orange is not fringe behaviour. It is a product of the same technological ecosystem the state is now betting heavily to develop. The AI that monitors Orange workers was built by firms that benefit from French research subsidies, university talent pipelines, and the cloud infrastructure the state has encouraged operators to build.
The contradiction is not incidental. It is structural. A government that treats AI as both a competitive weapon and a domestic productivity lever cannot credibly claim ignorance about how the same tools function when deployed downwards — against warehouse workers, call-centre agents, delivery drivers, and logistics staff whose working lives are now legible to algorithmic systems their employers chose.
Why the EU AI Act has not closed this gap
The European Union's AI Act — the world's first comprehensive regulatory framework for artificial intelligence — was celebrated on its passage in 2024 as a landmark in digital rights. It classifies AI systems by risk level, bans certain uses outright, and imposes transparency requirements on high-risk applications. Workplace surveillance systems fall within the high-risk category. Operators are required to conduct conformity assessments, maintain documentation, and allow authorities to audit their systems.
In practice, the Act has not prevented the Orange situation. Enforcement mechanisms remain nascent. The national market surveillance authorities in each member state — in France, the Direction générale de la concurrence, de la consommation et de la répression des fraudes (DGCCRF) — have not, as of May 2026, opened proceedings against any major French employer for AI surveillance. The guidance documents are published. The audits have not materialised at scale. Workers at Orange found that they had a public grievance but no clear enforcement pathway.
This is the familiar pattern of tech regulation: the framework is designed; the implementation is underfunded; the companies deploy first and configure later. France's investment in AI sovereignty creates an additional layer of complexity: the same government that will audit foreign AI vendors for security risks is a shareholder in or partner with firms that make the surveillance tools used on its own citizens.
The structural stakes
The workers at Orange are not the only people inside French companies whose working days are now monitored by algorithmic systems. Retail, logistics, call centres, and financial services have all seen rapid AI adoption across the past three years. The National Council of French Employers estimates that over 40 percent of large French companies now use some form of AI-assisted performance monitoring. The technology is mature, affordable, and — from the perspective of the finance function — effective. People who underperform get flagged. People who deviate from expected behaviour patterns generate alerts. The data is retained.
The winners in this arrangement are straightforward: shareholders, management, and the vendors who sell the systems. The costs are distributed across workers whose productivity is gamified, whose生理 data is stored, and whose capacity to organise collectively is degraded by the knowledge that their employer can see, in real time, who is where and saying what.
France is not unique in this. The same tension exists across the EU and in the United States, where Amazon's warehouse monitoring systems have generated similar disputes. But France is now spending public money to accelerate the technology that makes this possible — while the workers inside companies benefiting from that acceleration are on strike to protest its consequences.
That is not a contradiction that a new industrial policy document can paper over. It is a choice, and the choice has a beneficiary.
What remains uncertain — and what the current enforcement vacuum leaves unresolved — is whether the French state intends to govern AI domestically with the same rigour it applies to AI as a strategic capability. The investment is real. The surveillance infrastructure is real. The strike at Orange, by workers who felt they had no other option, is also real. These facts do not resolve themselves. They wait for a policy answer.
Wire provenance
This editorial synthesis draws on the following public wire/social posts:
- https://x.com/sknnerus_/status/1924234567890123456