AI at work: French courts keep suspending rollouts over missing works-council consultation — what it means for your SMB
On May 21, 2026, the Paris Court of Appeal confirmed the suspension of an internal AI writing assistant and company-wide ChatGPT use for lack of prior works-council (CSE) consultation, calling it an “manifestly unlawful disturbance.” Two more 2026 rulings build the same line. Here is what actually triggers the obligation, and the checklist before you roll out an AI tool.
On May 21, 2026, the Paris Court of Appeal confirmed the suspension of two AI tools deployed by Groupe Moniteur, a subsidiary of Infopro Digital: an internal writing assistant named “DIGI” and the company-wide rollout of ChatGPT. The reason: the employer had never opened an information-consultation procedure with its works council (comité social et économique, or CSE). The court ruled that this failure constitutes a “trouble manifestement illicite” — a legal qualification that lets a judge in summary proceedings order an immediate suspension without even having to prove urgency. The company was also ordered to pay €5,000 in damages to the CSE, for the anxiety the project caused among staff.
An old obligation the courts are now enforcing hard
The underlying obligation is nothing new: Article L. 2312-8 of the French Labor Code has long required employers to inform and consult the works council before any “significant project introducing new technology” likely to affect jobs, qualifications, pay, training or working conditions. What changed in 2026 is that courts started applying that text head-on to generative AI. The Nanterre court opened the sequence on January 29 by suspending the rollout of two AI-enabled HR software packages for lack of central-CSE consultation, then repeated the move on May 13 by ordering a company to open consultation on an internal tool giving access to several generative AI models. The Paris Court of Appeal’s May 21 ruling confirms that line and extends it explicitly to intellectual-production assistants — not just HR software.
What actually triggers the obligation — and what doesn’t erase it
- →It doesn’t matter whether the tool is free or consumer-grade: ChatGPT, Claude or any general-purpose assistant triggers the same obligation as soon as the employer generalizes, frames or facilitates its use.
- →It doesn’t matter what stage it’s at: a rollout labeled a “pilot” or an “experiment” still requires consultation once it involves real employees doing real tasks.
- →It doesn’t matter that some employees were already using the tool informally before the project: the obligation arises when the company officializes the use, not when the first employee opened the tool.
- →The scope goes well beyond customer-facing tools: a reporting agent, an email-triage assistant or an internal monitoring tool can just as easily qualify as a “significant project” if it changes a team’s working methods or workload.
What this means for your SMB
The trigger is the CSE threshold itself: from 11 employees up, the obligation applies — including in companies of 11 to 49 employees, where the CSE has reduced attributions but information-consultation on new technology is not among the exceptions. In practice, an SMB planning to give its customer service an AI support agent, hand its accounting team an automated overdue-invoice follow-up tool, or put an ops team under a monitoring watchdog that tracks indicators continuously must, once a CSE exists, present the project to elected staff representatives before the rollout — not after. This has nothing to do with EU AI Act compliance, whose deadlines cover different obligations: it is an older requirement under French labor law, and courts are now enforcing it with concrete consequences — suspension under a daily penalty, damages, legal costs, and a project delayed by months while consultation runs properly.
Before rolling out an AI tool: four checks
- →Take inventory of AI tools already in use across the company, including informal use by individual employees — that is often where the risk hides, not in the official flagship project.
- →If a CSE exists, open the information-consultation procedure before any company-wide rollout, even for a ChatGPT or Claude subscription presented as a simple productivity tool.
- →Prepare a file that sets out the tools involved, their precise functionality, the roles and tasks affected, and the expected effects on workload, training and working conditions — a vague file is the first thing a works council will contest in court.
- →Track this kind of case law with a regulatory watch agent rather than occasional trade-press reading: the case law shifted three times in five months, and that is exactly the sort of movement an SMB without in-house counsel tends to miss.
None of this is a reason to give up on AI — it is a reminder that rolling out an agent, however well built, also needs to be prepared socially, not just technically.
Frequently asked questions
Does a 15-employee SMB have to consult its works council before deploying ChatGPT or an AI agent internally?+
Yes. Once a company crosses the 11-employee threshold and has a works council (CSE), even with reduced attributions, Article L. 2312-8 of the Labor Code requires information-consultation before any significant project introducing new technology that affects working conditions — and 2026 case law confirms that generative AI, including a consumer tool like ChatGPT, falls within that scope.
What is a “manifestly unlawful disturbance” and why does that qualification matter?+
It is a legal qualification that lets a judge in summary proceedings order a measure — here, suspending an AI tool — without having to prove urgency. The Paris Court of Appeal applied it on May 21, 2026 to a failure to consult the works council, which considerably speeds up the procedure for staff representatives.
What are the concrete consequences of failing to consult?+
A judge can order the immediate suspension of the tool under a daily penalty until consultation is completed, order the company to pay damages — €5,000 in the May 21, 2026 case, for the works council’s anxiety-related harm — and cover legal costs. The rollout is then delayed by several additional months.
Does this obligation only cover customer-facing tools?+
No. The 2026 case law concerns internal tools — a writing assistant, HR software, access to generative AI models — as soon as they affect employees’ working methods, workload or skills. An internal reporting or email-triage agent can be just as concerned as a customer-facing chatbot.
Free resource
The self-assessment grid: 20 tasks AI can automate
Sales, admin, support, operations: the 20 tasks AI agents already handle in SMEs — with, for each one, the tell-tale sign that your team is concerned.
Read next
Business & compliance
EU AI Act: what became mandatory on August 2, 2026 — and what it means for your SMB
19 August 2026·5 min read
Business & compliance
Claude now watermarks its text: what Anthropic’s move changes — and doesn’t — for your SMB’s AI content
18 August 2026·5 min read
Business & compliance
Computer History: ChatGPT now remembers your activity — except in France (for now)
17 August 2026·5 min read