Ethics in Artificial Intelligence for Food and Health: From “Can Do” To “Should Do”

CLARA TALENS, researcher at New Foods

Artificial intelligence (AI) is already present throughout the entire agri-food chain: from predicting food safety risks to designing personalised diets and supporting clinical diagnosis. This leap brings benefits, efficiency gains, new evidence and faster decision-making, but also risks: biased data, opaque models, unsupervised use and security gaps that can directly affect public health. AI ethics is not a brake: it is the operational framework that ensures innovations work for everyone and build trust.

Why Ethics Matters (Especially) in Food and Health

In the food sector, artificial intelligence offers unique opportunities: optimising traceability, reducing food waste, improving nutritional quality or anticipating food safety risks. However, if algorithms are trained on incomplete or poorly representative data, they can reinforce inequalities and undermine consumer trust. International bodies such as FAO highlight this double edge: AI can boost the sustainability and climate resilience of the agri-food system, but it can also entrench asymmetries if it is applied without transparency, without involving vulnerable stakeholders (for example, small producers or rural communities) or without proper data governance.

A relevant case in the field of personalised nutrition shows how an unethical or technically deficient use of AI can lead to wrong decisions. A study published in Nutrition & Diabetes (Nature, 2022) warns that machine learning models used in nutritional interventions can produce misleading dietary recommendations when they are trained on incomplete or biased data – for example, diets based on Western patterns that ignore the cultural and metabolic diversity of different populations. The authors stress the need to design and validate these systems in an iterative and ethical way, to prevent automated recommendations from worsening nutritional inequalities instead of reducing them.

What Current Policies and Standards Require

In the areas of food and health, regulation and international guidelines are moving towards a model of responsible, safe and traceable artificial intelligence that guarantees both consumer protection and the scientific reliability of food systems.

FAO and WHO (AI in food and health, 2021–2025)
Both organisations recommend that the use of AI in nutrition, production and food safety should incorporate data transparency, traceability of automated decisions and human oversight. They emphasise that any system affecting the availability or quality of food must be assessed not only for its efficiency, but also for its ethical and social impact, avoiding the widening of inequalities in access to healthy and sustainable diets.

European Artificial Intelligence Act (AI Act)
Now in force, the AI Act classifies as high-risk those systems that influence human safety or health, including AI applications in quality control, nutrition labelling, food risk assessment or dietary advice. It requires documented risk management, the use of high-quality and representative data, clear information for users and effective human oversight. This marks a structural shift: ethics ceases to be a recommendation and becomes a regulatory obligation.

ISO/IEC 23894:2023
This international standard sets out a risk management framework for AI that can also be applied in the food industry, aimed at identifying, analysing, treating and monitoring technical, ethical and organisational risks. In the agri-food context, its adoption makes it possible to integrate ethics from the design phase of AI systems for production, traceability, contaminant control or the formulation of new foods.

EFSA (European Food Safety Authority)
The European food safety authority is promoting the use of AI to strengthen “data readiness” in risk assessment, automate scientific reviews and detect early signals of contamination or fraud. However, it warns that AI will only be useful if it is based on auditable, well-governed and accessible data, ensuring reproducibility and trust in regulatory decisions.

A Practical Framework for Using AI Ethically in Food and Health

Before developing or implementing AI systems in food and health, it is useful to ask a few key questions to guide responsible design and decision-making:

1. What outcome are we aiming for: improving public health, sustainability or food equity – and what potential risks does the system entail? Is it classified as “high risk” under the AI Act?
2. Where do the data come from? Do they adequately represent different sexes, ages, regions and socioeconomic levels? Have biases or limitations been identified?
3. Does the model work equally well for all groups? What happens if only one variable changes, such as sex or postcode?
4. Can we understand why the AI system proposes a particular decision or dietary recommendation? Is there clear documentation and version control for the model?
5. Is there human oversight of critical decisions? Is model drift monitored and are potential biases audited after deployment?
6. Have technologists, health professionals, quality managers and consumers been involved in the process? Are impacts and lessons learned communicated transparently?

What’s Next: Regulation and Trust

The gradual implementation of the AI Act in Europe will mark a turning point in the responsible use of artificial intelligence, including in the food and health sectors, where data sensitivity and impacts on people are especially significant. This new regulation will require greater process traceability, tighter control over data quality and provenance, model explainability and validation in real-world settings before deployment. Anticipating these requirements will not only reduce regulatory risks, but also strengthen the trust of consumers, authorities and businesses – a critical asset if AI is to be integrated sustainably into the food industry.

Applied AI ethics is also a competitive advantage for the sector. Taking an ethical approach from the design stage – with clear purposes, risk management, representative and transparent data, explainable model decisions and bias audits – makes it possible to turn information into reliable, traceable and socially acceptable decisions. Ultimately, ethical AI not only accelerates innovation: it also improves public health, reinforces food safety and supports a more sustainable and fair food system for all people and communities.

Key references

• Thomas, D. M., Kleinberg, S., Brown, A. W., et al. (2022). Machine learning modeling practices to support the principles of AI and ethics in nutrition research. Nutrition & Diabetes, 12(48). https://doi.org/10.1038/s41387-022-00226-y
• European Commission. (n.d.). AI Act: Regulatory framework for artificial intelligence. Retrieved from https://digital-strategy.ec.europa.eu/en/policies/regulatory-framework-ai
• International Organization for Standardization. (2023). ISO/IEC 23894:2023. Artificial Intelligence—Guidance on risk management. Geneva: ISO.
• European Food Safety Authority (EFSA). (n.d.). Aplicaciones de IA en evaluación de riesgos y data readiness. Retrieved from https://www.efsa.europa.eu/
• Food and Agriculture Organization of the United Nations (FAO). (n.d.). IA en agrifood: oportunidades, riesgos y consideraciones éticas. Retrieved from https://www.fao.org/