LEGO flags illegal job offers before France Travail publishes them

The Cour des comptes found that France Travail's LEGO model catches illegal job offers before publication and cut advisers' checking work by the equivalent of 98 full-time posts.

  • 2 an organization

    By the operator's figures, flags became correct 93% of the time against 65% before, and checking work fell by the equivalent of 98 full-time posts.

  • Not rated society

    Stopping illegal offers keeps discriminatory job ads off the site, but the Cour gives no count of illegal offers stopped.

Depth runs from 1, a convenience or time saved, to 5, a life or a livelihood preserved.

Status
Reported benefit.
Evidence

foundstudiedreportedclaimed

An independent audit report supports the benefit.

Compared with

TextKernel, the keyword software used until autumn 2021, run in parallel during 2021. Measured.

Without the system
« seulement » 65 % avec l’ancien logiciel
With it
93 % des offres d’emploi signalées par LEGO le sont à raison
The system’s part
Not isolated. During 2021 LEGO and the old keyword software checked the same offers in parallel, which compares the two systems. The figures come from France Travail, and the Cour compares no checking without software. The Cour states that LEGO uses natural language processing by machine learning with deep neural networks.
Limits
Not yet checked.
Costs and harms
Not yet checked.
Later
Not yet checked.

Download data Cite Suggest a correction

Benefit type
Improved Task Accuracy
System
LEGO (Légalité des offres), natural language processing with deep neural networks
Made by
Pôle emploi, now France Travail
Used by
France Travail (business advisers in all agencies)
Where
France (national)
Stage
Expanding
Date
, when the source was published
Kind
Aggregate result
Broad benefit
Evaluating Information to Determine Compliance with Standards
Collections
Scrutiny and findings
Last revised
31 October 2026

Sources

  1. Audit report, by authors with no tie to the system.

    Cour des comptes. France Travail et l’intelligence artificielle. Observations définitives S2025-1558, adopted 15 October 2025 and published 8 January 2026, p. 9 and annex 5, pp. 111-112.

    Certains, déployés dès le programme Intelligence emploi, ont démontré leur efficacité (analyse automatique de CV, identification des offres d’emplois illégales, Lego, etc.).

Details

AI system

Workflow

The model checked each offer against 23 legality criteria before publication. Business advisers handled the flagged offers each day by approving or rewording them.

Follow-up

Conditions for replication

Advisers must review each flagged offer before it is published or changed.

Entry details

Sector

Public employment services.

Purpose

Pôle emploi built LEGO to detect possibly illegal job offers before publication and to alert the advisers who check them.

Not yet checked: model, release, provider, role, configuration, inputs and outputs, autonomy, human oversight, AI contribution, beneficiaries, reach, distribution and access, duration, every measurement question, every money question, ownership and control, dependencies, cross-border connections, geopolitical implications, replication attempts, availability, institutional decision, grounds, action and result, success criterion, every evidence question, observed or intended, alternative assessments, related entries, other identifiers, category mappings.

Cite this record

Peter H. Vartanian. (2026). LEGO flags illegal job offers before France Travail publishes them (AIBD-0683). Artificial Intelligence Benefits Database, edition of 31 October 2026. https://petervartanian.xyz/aibd/records/AIBD-0683.html (archived at https://doi.org/10.7910/DVN/AADSII, version 4.0)

BibTeX Data Field definitions