NewsCore
    veille anticorruptionconformité Sapin IIévaluation des tierscartographie des risquesdue diligenceveille presse conformité

    Anti-corruption monitoring: turning the media signal into Sapin II compliance evidence

    Effective anti-corruption monitoring is not about reading the press, but about linking each signal to a Sapin II obligation. Method, governance and the role of OSINT technology for compliance departments.

    17 August 20267 min read

    The Sapin II law requires companies with at least 500 employees and more than 100 million euros in revenue (thresholds set out by the Agence française anticorruption, AFA) to operate a complete anti-corruption programme, built around an up-to-date risk map and continuous third-party assessment. Yet a large share of the programmes audited by the AFA have a risk map judged incomplete or too static: according to the law firm Novlaw Avocats, which analyses feedback from these audits, the agency particularly regrets risk scenarios that are too generic and transferable from one group to another. The breaking point is rarely the will to comply: it is the ability to capture, at the right moment, the media signal that reveals a risk at a supplier, a sales agent or an acquisition target. This article shows how structured anti-corruption monitoring turns a raw news flow into evidence that stands up before an auditor.

    The essentials in 30 seconds
    • Anti-corruption monitoring is not about reading the press, but about linking each media signal to a specific Sapin II obligation: risk mapping, third-party assessment, accounting controls.
    • AFA audits look as much at the real effectiveness of measures as at their existence: untracked monitoring is hard to defend.
    • A tooled-up system cuts the time between the publication of an adverse article and its inclusion in the compliance file from several weeks to a few hours.

    What Sapin II really requires in terms of third-party monitoring

    Article 17 of the law of 9 December 2016 structures the anti-corruption programme around 8 pillars. Three of them depend directly on reliable, continuous external information: risk mapping, third-party assessment procedures and the accounting control system. On these three pillars, the most frequent failing is not the absence of a written procedure, but the inability to demonstrate that monitoring actually worked over time.

    In practice, an auditor does not merely check that a policy exists. They ask for proof that a supplier classified as high risk was indeed subject to enhanced monitoring, and that press alerts concerning it were tracked, qualified and, where appropriate, escalated. The AFA itself acknowledges, in its guide on third-party assessment, that this system continues to raise practical difficulties for many companies. Anti-corruption monitoring then becomes the documentary backbone of the programme: without it, the risk map remains a snapshot frozen at the date it was drawn up.

    Press monitoring, the blind spot of anti-corruption programmes

    Most compliance departments have transactional screening tools (sanctions lists, PEPs, company databases). But these databases are structurally behind: a company rarely appears on an official list before a judicial investigation has been opened. Yet in many corruption cases, the first public signal comes from the media. A press investigation, a prosecutor's statement, an NGO report or a regional business article often precede inclusion in a database by several months.

    Setting up automated compliance monitoring on the scope of high-risk third parties makes it possible to close this time gap. The aim is not to monitor everyone, but to focus detection on the third parties identified as sensitive by the risk map: sales intermediaries, suppliers in exposed jurisdictions, joint venture partners. Well-calibrated monitoring does not generate noise: it produces a limited number of qualified alerts, each attributable to a third party and a risk level.

    Methodology: link each signal to an obligation

    Robust anti-corruption monitoring rests on a four-stage processing chain. First, defining the scope: the list of third parties drawn from the risk map, enriched with their executives and beneficial owners. Next, multilingual collection, because a corruption case involving a subsidiary is often first covered by the local press. Then qualification: each article is attached to a third party, a risk type (bribery of a public official, conflict of interest, influence peddling) and a severity level. Finally, documented escalation to the compliance officer.

    Traceability as evidence

    The decisive reflex is to timestamp every step. A signal detected on Tuesday, qualified on Wednesday and escalated on Thursday tells a story that stands up. For firms that outsource part of this workload, intelligence for compliance consultancies brings the documentary rigour the AFA expects while lightening the operational load on internal teams. Compliance evidence is not manufactured after the fact: it is built continuously, signal by signal.

    From collection to evidence: the role of OSINT technology

    Volume makes manual monitoring illusory. A portfolio of a few hundred third parties, multiplied by name variants, subsidiaries and languages, generates a flow that no one can read in full. This is where proprietary OSINT technology comes in: it does not just surface articles containing a name, it disambiguates entities, rules out homonyms and prioritises signals according to their severity and source reliability.

    An effective system must also reason in terms of entities, not keywords. Detecting that an executive of a supplier is named in an investigation, even if the company's name does not appear in the article, is precisely the kind of correlation a human misses under the pressure of volume and that a semantic analysis engine delivers systematically.

    Example: a local press article about a sales intermediary

    Take a French industrial group that relies, for its sales in a West African country, on a sales agent classified as high risk by its risk map. One Tuesday morning, a regional daily in that country publishes, in the local language, an investigation naming the agent's director in a case involving public contracts awarded to associates of a ministry official. The agent's own name does not appear in the article.

    In a manual system, the information is very likely to go unnoticed: it is not in French, not in a major outlet, and not linked to the name of the company being monitored. In a tooled-up system, the engine recognises the director as the beneficial owner of the third party, translates the article, classifies it under the type "bribery of a public official" with a high severity level, and sends it to the relevant compliance officer.

    The officer then has a complete trail: publication date, detection date, qualification, decision taken (request for explanations from the agent, suspension of payments, review of the relationship). This is exactly what an auditor will ask for: not proof that the company reads the press, but proof that it reacted, within a reasonable time, to a signal concerning a third party identified as sensitive.

    The same case illustrates the value of linking this monitoring to other systems: the information feeds the reassessment of the risk map, adds to the third party's due diligence file and can switch to a crisis monitoring logic if the group's name ends up being publicly associated with the case.

    Industrialising monitoring without overloading teams

    The risk of poorly designed monitoring is well known: alert fatigue. Too many unqualified notifications, and teams end up ignoring the flow, which destroys the evidential value of the system. The right approach is to set severity thresholds, route automatically to the right officer and reserve human intervention for the signals that warrant it. The table below contrasts ad hoc monitoring with industrialised monitoring on the NewsCore platform.

    DimensionManual monitoringTooled-up monitoring
    Detection timeSeveral weeksA few hours
    Language coverageLimited to FrenchMultilingual by default
    TraceabilityFragmented (emails, files)Timestamped and centralised
    Homonym riskHighDisambiguated by entity

    Anti-corruption monitoring is not an extra layer of reporting: it is the mechanism that keeps your risk map alive and your programme defensible. To go further on automating compliance screening, see our dedicated guide to KYC and sanctions screening, a natural complement to robust anti-corruption monitoring.

    Frequently asked questions

    Is anti-corruption monitoring mandatory under Sapin II?

    The law does not explicitly require a monitoring tool, but it does require third-party assessment and an up-to-date risk map. In practice, demonstrating continuous monitoring of high-risk third parties without a structured monitoring system is very difficult in an AFA audit.

    Should all suppliers be monitored?

    No. The risk-based approach recommends focusing enhanced monitoring on the third parties classified as high risk by the risk map: intermediaries, exposed jurisdictions, sensitive sectors. Undifferentiated monitoring dilutes the effort and drowns out important signals. Low-risk third parties fall under periodic review.

    How do you avoid alert fatigue?

    By calibrating severity thresholds, disambiguating entities to eliminate false positives and routing each alert to the right officer. Monitoring that produces few alerts, all of them relevant, is far more effective than a massive, unqualified flow.

    Which sources should be monitored first for anti-corruption monitoring?

    Three families: the local and business press of the countries where high-risk third parties operate, statements from judicial and regulatory authorities, and reports from specialised NGOs. This scope largely overlaps with that of ESG monitoring and regulatory monitoring, which makes it possible to pool collection and qualification between compliance and CSR departments.

    Sources

    Ludovic Desgranges, CEO NewsCore

    Go deeper

    All reports

    Three NewsCore reports that build on this article.