Food safety breakthrough: AI tool flags risky suppliers using global data metrics

The framework begins by establishing a standardized process to score suppliers based on seven risk indicators: hazard risk, commodity vulnerability, incident category, audit performance, logistic performance index (LPI), GDP per capita, and GDP growth. Each metric is evaluated on a normalized scale and then combined into an overall risk score (ORS) using customizable weights, which reflect the relative importance of each criterion.

The global food supply chain is becoming increasingly complex, demanding enhanced oversight mechanisms to keep consumers safe. In a newly published study titled “Supplier Risk Assessment—A Quantitative Tool for the Identification of Reliable Suppliers to Enhance Food Safety Across the Supply Chain” in Foods, researchers Sina Röhrs, Sascha Rohn, Yvonne Pfeifer, and Anna Romanova unveil a comprehensive, AI-ready framework designed to revolutionize how food industry stakeholders assess supplier reliability and mitigate contamination risks.

With over 600 million people falling ill annually due to contaminated food, ensuring supplier compliance has become a global imperative. The proposed tool, developed through SGS’s Digicomply platform, introduces a quantifiable, multi-factor assessment model capable of automatically scoring supplier risk based on a wide array of inputs. The framework integrates real-world data, risk normalization techniques, and customizable weightings, bringing both structure and adaptability to supplier vetting practices.

How does the model assess supplier risk in real-world food systems?

The framework begins by establishing a standardized process to score suppliers based on seven risk indicators: hazard risk, commodity vulnerability, incident category, audit performance, logistic performance index (LPI), GDP per capita, and GDP growth. Each metric is evaluated on a normalized scale and then combined into an overall risk score (ORS) using customizable weights, which reflect the relative importance of each criterion.

The model’s core strength lies in its dynamic structure. It assesses historical data on food safety incidents using a weighted matrix that prioritizes consumer risk. It incorporates economic metrics such as GDP and logistics efficiency to contextualize supplier environments and detect structural vulnerabilities. Commodity vulnerability is assessed through a 12-month rolling analysis of incident reports, while incident categories, ranging from recalls to consumer complaints, are assigned severity scores.

Importantly, the assessment design allows for AI integration. Data sourced from public databases like RASFF and other AI-compatible systems can feed directly into the risk matrix, allowing near real-time risk updates. For example, a supplier with recent recalls or whose country shows signs of economic instability would automatically see their risk profile escalate. This dynamic scoring capacity makes the tool particularly powerful in fast-changing supply chain environments.

What makes this approach different from traditional supplier audits?

Traditional supplier evaluation methods rely heavily on static audits and manual review processes. These are time-consuming, expensive, and often outdated by the time results are implemented. In contrast, the AI-compatible framework presented in the study automates much of the risk evaluation process, providing both speed and consistency.

Audit performance remains a key indicator within the model, but the framework addresses a known limitation: lack of access to audit data. Suppliers often restrict audit transparency for competitive or reputational reasons. To address this, the authors propose flexible weighting, allowing the system to redistribute the significance of other metrics when audit data is unavailable.

This modularity makes the tool broadly usable across different market scenarios. In high-transparency contexts, audits can play a central role. In opaque markets, the model leans more heavily on economic, logistical, and incident data. This flexibility ensures applicability across the full spectrum of food industry stakeholders - from multinational retailers to small-scale processors.

Further, the tool doesn’t just detect risk; it pressures suppliers to improve. Public availability or customer-facing integration of these scores could incentivize better practices, driving transparency and raising baseline standards. Suppliers who know they’re being continuously scored are more likely to invest in compliance, traceability, and quality assurance, especially when risk ratings influence procurement decisions.

What are the broader implications and future directions for food safety management?

The researchers stress that this tool is not merely conceptual, it was validated through manual application across 11 real-world suppliers. These ranged from beverage producers implicated in food fraud to ice cream vendors with labeling issues. The risk scores calculated by the system aligned closely with known safety records, highlighting its practical viability. For instance, suppliers associated with E. coli outbreaks, fraudulent vodka production, or allergen mislabeling consistently scored in the moderate- to high-risk range.

However, the tool’s greatest potential lies in what comes next: full-scale integration into AI databases. Platforms like Digicomply already manage millions of food safety entries across over 160 countries. Embedding the supplier risk model into such databases would allow predictive analytics at scale. The system could flag emerging risks, suggest alternate vendors, and even simulate risk under different procurement scenarios.

Still, challenges remain. Audit data transparency is a major barrier, as are disparities in regulatory rigor across countries. Economic indicators like GDP growth, while helpful, must be interpreted contextually. A low-growth nation with historically robust infrastructure may pose less risk than a high-growth country with poor food safety enforcement.

Additionally, the authors propose expanding the model to include environmental data, such as weather conditions or sustainability metrics. Climate change-induced shifts in temperature and humidity can trigger contamination events like mycotoxin growth - risks not currently captured in most supplier evaluations. Similarly, integrating sustainability indicators like carbon emissions could help align procurement with ESG goals, making food systems not just safer, but more responsible.

To ensure lasting impact, the study suggests moving beyond voluntary frameworks. Regulatory bodies could consider standardizing supplier risk assessments through AI-driven models, creating a shared language for food safety risk that transcends borders. Harmonized risk scoring would enable mutual recognition between countries and reduce the compliance burden on exporters and importers alike.