Larger Datasets Could Improve Predictive Food Safety

Participants consistently said that pooling food safety data could provide larger, more diverse datasets, allowing companies to identify trends and root causes that might not be apparent from an individual company's data. Participants also saw potential for larger datasets to support predictive modeling and earlier identification of food safety risks.

Smaller companies could especially benefit from pooled insights because they may lack the resources to independently conduct extensive research and data analysis, according to interviewees.

The researchers noted that these larger, more comprehensive datasets could support artificial intelligence (AI) and machine learning applications for food safety, provided the aggregated data are accurate, unbiased, and representative of their intended applications.

However, technical barriers would need to be overcome. Participants reported uneven levels of digitalization across the industry, with some companies still relying on paper records, spreadsheets, or other systems that complicate data aggregation. Lack of standardized data collection practices, terminology, formats, and protocols was also identified as a significant obstacle.

Trust, Liability Concerns Impede Food Safety Data Sharing

Trust emerged as an overarching factor influencing companies' willingness to share confidential food safety data.

Participants expressed concerns that once information left their control, it could be misinterpreted, taken out of context, accessed by unintended parties, or used against the company. Specific concerns included legal liability, regulatory scrutiny, reputational damage, loss of customers, and the possibility that competitors could use shared information to gain an advantage.

These concerns created what the researchers described as a collective action problem: Although shared data could produce broader food safety and public health benefits, individual companies would bear much of the cost and risk associated with participating.

The study also highlighted how differences in food safety programs could complicate interpretation. For example, a company with a rigorous environmental monitoring program could report more positive pathogen findings than a company conducting less extensive testing. Without adequate context, the higher number of positives could incorrectly be interpreted as evidence of poorer food safety performance.

Data Governance Could Help Build Trust

Participants identified data governance as a potential means of addressing trust-related concerns. Suggested approaches included clear rules governing data ownership, access, confidentiality, retention, and use, as well as neutral third-party oversight.

Academic institutions were specifically suggested as possible intermediaries because they could facilitate data sharing while helping protect confidentiality and prevent competitors from directly accessing or misusing companies' information.

Participants also emphasized that data-sharing initiatives should have clearly defined objectives rather than pooling information without a specific food safety problem to address.

Based on the findings, the researchers recommended piloting a governance framework for voluntary horizontal food safety data sharing involving industry, regulatory bodies, and academic institutions. They also suggested that smaller, more homogeneous groups of companies, such as firms with similar geographic locations or product risk profiles, could provide a more practical starting point than an industry-wide approach.

Companies Weigh Risks vs. Benefits

Overall, the researchers concluded that limited horizontal food safety data sharing did not simply reflect companies' unwillingness to collaborate. Rather, companies were weighing collective food safety benefits against individualized legal, reputational, competitive, and financial risks. Addressing these risks through effective data governance could therefore be important to enabling broader collaborative data analysis and realizing the potential of AI-driven food safety applications.

The study was conducted by researchers at Cornell University, the University of California–Davis, and the University of California–Berkeley.