The **AI Act**, a proposed European regulation on artificial intelligence, has been under discussion among leaders, with potential penalties for non-compliance. A meeting to address high-risk AI systems is planned for 2027, but in the meantime, two French companies have been forced to stop using their AI tools. This was not ordered by the European Union but by a French judge handling urgent matters, based on the Labor Code. The difference between the theoretical risk of AI and the actual risk being managed is significant. Two subsidiaries of a major professional press group had introduced a writing assistant they developed internally and also allowed the use of ChatGPT through an IT policy. Neither company consulted their social and economic committee (CSE) on these changes. On July 15, 2025, the judicial court of Créteil ordered the suspension of these AI tools until the consultation was completed, with a provisional daily fine of 1,000 euros for three months and a deposit of 5,000 euros. These decisions were upheld by the Paris Court of Appeal in May 2026, adding 3,000 euros per file under article 700 of the Civil Procedure Code. The legal basis for the court's decision was article L2312-8, II, 4° of the Labor Code, which requires consultation with the CSE when new technologies are introduced and could affect employment, qualifications, pay, training, or working conditions. The failure to consult was deemed a "manifestly illegal disturbance" under article 835 of the Civil Procedure Code. The Paris Court of Appeal confirmed this in its rulings on May 21, 2026. The full text of the ruling is available for review. In France, administrative fines typically require a formal authority, a procedure, and an investigation. However, the law establishing the control authorities for the **AI Act** had not been adopted by mid-September 2026. While the risk of non-compliance is serious, it remains delayed. Urgent court orders, on the other hand, are immediate and not subject to negotiation. These orders do not directly impose financial penalties but instead require the immediate suspension of AI tools. For a company that has integrated an AI assistant into its workflow, this means a sudden stoppage of operations rather than a financial hit. A manager can prepare for a fine, but not for an unexpected halt in processes. Three key points are often overlooked. First, article L2312-8 applies only to companies with at least fifty employees. For companies with fewer than fifty, the relevant articles (L2312-5 to L2312-7) do not require prior consultation on new technologies. Clarifying this helps avoid unnecessary concern and reinforces the legitimacy of the rules. Second, a negative opinion from the CSE does not block the implementation of AI tools. The CSE provides an opinion, not a veto, and if no opinion is given within the required timeframe, it is considered negative. The issue is not the CSE’s opposition, but the failure to consult. Third, the consultation must occur before the deployment of AI tools, not after. In both cases, the tools were already in use, which led to the court's ruling. The legal implications of AI use extend beyond the CSE consultation. On September 15, 2026, the same court suspended a project at one of the companies to eliminate eight editorial secretarial positions, imposing a daily fine of 8,000 euros. The management reserved the right to appeal. This decision was based on a different legal basis: a serious risk to the physical and mental health of employees, supported by a risk assessment that highlighted the absence of preventive measures. Nineteen journalists were affected, and twenty-six publications were impacted. This shows that the relationship between AI and labor law is not limited to CSE consultations but also includes risk assessments, which can be just as immediate in their effects. For managers, one practical and cost-free step is to inventory and survey all AI tools currently in use, including those that have not been formally adopted. They can then draft clear rules—just a few pages—specifying which AI tools are authorized, for what purposes, and what data should never be used. These rules should also define who is responsible for approving new uses. This approach aligns with article 4 of the European AI regulation, which came into effect in February 2025 and was updated in July 2026. This article requires companies using AI to promote AI mastery among their employees, without a minimum employee threshold. The key is to consult before implementing AI tools, not after, and to keep a detailed record of all actions. A measure that cannot be proven will be difficult to enforce. The author is not a lawyer, and this article is based on public legal decisions and does not provide legal advice.