How can AI monitor supplier risk and compliance in fashion?
Forced labour rules and due diligence laws push fashion companies to know their suppliers better. AI can scan documents and signals at scale, but it cannot replace audits and engagement.
KEY TAKEAWAYS Summary by the editors
- AI supplier risk monitoring uses machine learning and language models to screen supplier data, documents, news and trade signals for labour, environmental, financial and disruption risks.
- The EU Forced Labour Regulation (EU) 2024/3015 applies from 14 December 2027 and covers products of all sectors, at any stage of production.
- Commission guidelines published on 30 June 2026 state that robust due diligence is not a safe harbour against product bans under the Forced Labour Regulation.
- The amended Corporate Sustainability Due Diligence Directive now covers companies with at least 5,000 employees and €1.5 billion turnover, with obligations applying from 26 July 2029.
- AI can prioritise where to look, but findings depend on knowing lower-tier suppliers and must be verified through audits, documents and worker engagement.
AI can monitor supplier risk by screening large volumes of supplier data, documents, news and trade information for signs of labour abuses, environmental problems, financial distress or disruption. It helps compliance and sourcing teams decide where to look first. It cannot prove that a supply chain is clean, and its signals must be checked through audits, documentation and engagement with suppliers and workers.
Why is supplier risk a growing priority for fashion?
Fashion supply chains are long and fragmented. A garment may involve a cotton farm, a spinner, a weaver, a dye house and a cut and sew factory in different countries, and many brands only have direct relationships with the last step. Risks in the deeper tiers, from forced labour to pollution, are harder to see but increasingly regulated.
Regulation (EU) 2024/3015 prohibits products made with forced labour on the EU market. According to legal analysis by Hogan Lovells Cadwalader, it applies from 14 December 2027 to all products placed or made available on the EU market or exported from it, regardless of origin, sector or stage of the value chain. Companies are expected to consult the Commission's Forced Labour Risk Database when identifying where risks arise.
Guidelines published by the Commission on 30 June 2026 recommend risk-based due diligence along the OECD framework, while making clear, according to the same analysis, that robust due diligence is not a safe harbour against a product ban. The same summary notes that authorities may rely on sources such as civil society and trade union reports, customs and trade data, satellite imagery and scientific methods like isotopic testing.
What does AI do in supplier risk monitoring?
| Application | What AI does | Key limitation |
|---|---|---|
| Adverse media screening | Scans news and reports in many languages for supplier names and risk topics | Name matching errors; coverage gaps in some regions |
| Document review | Extracts and checks data from certificates, audit reports and contracts | Cannot detect falsified documents reliably |
| Supply chain mapping | Links suppliers, sub-suppliers and products from purchase and trade data | Only as complete as the underlying data |
| Risk scoring | Combines country, product and supplier indicators into priorities | Scores can hide assumptions and bias |
| Disruption alerts | Flags weather, logistics or financial signals affecting suppliers | Many alerts are noise without context |
Language models are particularly useful for reading unstructured material: audit reports in different formats, supplier questionnaires, certificates and news articles in local languages. Classification models can then group findings by risk type, such as the ILO-derived forced labour indicators referred to in the Commission guidelines, including deception, restriction of movement, debt bondage, withholding of wages and excessive overtime.
How does due diligence law shape AI use?
The Corporate Sustainability Due Diligence Directive was amended by Directive (EU) 2026/470, published in February 2026. According to Noerr's analysis, it now applies to companies with at least 5,000 employees and a global net turnover of at least €1.5 billion, the transposition deadline is 26 July 2028 and obligations apply from 26 July 2029. Companies may focus on the areas where adverse impacts are most likely and most severe.
That risk-based approach suits AI tools, which are good at prioritising. But it also places responsibility on companies to justify how they set priorities. In February 2026 the OECD published Due Diligence Guidance for Responsible AI, which applies responsible business conduct standards to companies that develop and use AI. Using AI in due diligence therefore brings its own duty to understand the tool's limits.
Which data sources feed AI supplier monitoring?
The quality of AI risk monitoring depends on its inputs. Internal sources include supplier master data, purchase orders, audit reports, certificates and corrective action plans. External sources include news, reports by civil society organisations and trade unions, sanctions and enforcement lists, customs and trade data and, for some materials, satellite imagery. The Commission guidelines on forced labour list several of these as sources that authorities may themselves rely on, which suggests companies should be aware of them too.
Combining these sources is where AI adds most value. A model can link a news report about a spinning mill to the fabric suppliers that buy from it, and from there to the styles a brand sells. That only works if the brand knows the relationships between its suppliers. For this reason, supplier mapping, often started for compliance, is also the precondition for useful AI monitoring.
Companies should also consider how risk signals are shared. Suppliers are more likely to cooperate if alerts lead to dialogue and remediation rather than immediate termination, and the OECD due diligence approach emphasises responsible engagement with business partners.
What are the limits of AI supplier monitoring?
- Unknown tiers: AI cannot assess suppliers that are not in the company's data.
- Hidden abuse: forced labour is often concealed and rarely shows up in news or documents.
- Bias in sources: regions with less media coverage can appear low-risk simply because less is reported.
- False positives: name confusion between companies can trigger unwarranted alarms and damage relationships.
- Data protection: processing information about individuals, such as workers or managers, must comply with privacy law.
How should a fashion company start?
- Map direct suppliers completely and begin collecting data on key lower-tier suppliers, starting with high-risk materials such as cotton.
- Define risk categories and escalation rules before choosing tools.
- Pilot AI screening on a subset of suppliers and compare its findings with existing audit results.
- Assign people to review every high-priority alert and record decisions.
- Share findings and expectations with suppliers, since remediation depends on cooperation.
Frequently asked questions
When does the EU forced labour ban apply?
Regulation (EU) 2024/3015 entered into force on 13 December 2024 and applies from 14 December 2027. It covers products of all sectors made with forced labour at any stage of the supply chain.
Can AI detect forced labour in supply chains?
AI can flag risk indicators in documents, news and trade data and help prioritise suppliers for review. It cannot prove the presence or absence of forced labour, which requires audits, documentation and engagement with workers.
Which companies does the CSDDD now cover?
Following Directive (EU) 2026/470, the CSDDD covers companies with at least 5,000 employees and €1.5 billion global net turnover, plus certain non-EU companies with more than €1.5 billion turnover in the EU. Obligations apply from 26 July 2029.
Is due diligence a defence against the EU forced labour ban?
According to the Commission guidelines of June 2026, as summarised by legal advisers, robust due diligence is not a safe harbour. A product can still be banned if forced labour is established, although due diligence helps companies identify and address risks earlier.
One edition every weekday morning. Read in five minutes. Free for industry professionals.
SOURCES
- EUR-Lex: Regulation (EU) 2024/3015 on prohibiting products made with forced labour
- Hogan Lovells Cadwalader: EU Forced Labour Regulation, key takeaways from the European Commission's June 2026 guidelines
- EUR-Lex: Directive (EU) 2026/470 amending the sustainability reporting and due diligence directives
- Noerr: CSDDD amending directive published in the Official Journal of the European Union
- OECD: OECD Due Diligence Guidance for Responsible AI
