Responsible Artificial Intelligence in Diabetes Care: The IDF Perspective at EASD 2026

Artificial intelligence is already part of diabetes care. Insulin pumps that adjust doses automatically, software that predicts low blood glucose and tools that screen for eye disease are in daily use, and conversational assistants are now answering questions from people living with diabetes. The question for governments is no longer whether these tools will enter health systems, but how they will be governed.

On 29 September 2026, Dr Amit Kumar Dey, Chair of the International Diabetes Federation (IDF) Technology and AI Working Group, addressed this question at EASD 2026 in Milan in a session titled “AI in Diabetes 2026: From Agentic Systems to Responsible Deployment”. He set out the scale of the challenge using the IDF Diabetes Atlas: 589 million adults were living with diabetes in 2024, a number projected to reach 853 million by 2050, and around 252 million were undiagnosed. More than three in four live in low- and middle-income countries, where specialists are fewest and, as Dr Dey noted, where the promise and the risk of AI are greatest.

Dr Dey described four levels of clinical AI, each carrying more risk than the last: tools that describe what has happened, tools that predict what will happen, tools that recommend an action, and tools that take the action themselves. He cautioned that the evidence is uneven. The only systematic review of diabetes self-management chatbots identified 25 studies, of which just one was a randomised controlled trial (Wu et al., 2024), and only 5 of 40 studies on predicting type 2 diabetes risk had been validated in populations beyond those used to build them. In his words, “Chatbots need RCTs before we make outcome claims.” 

He also presented his team's evaluation of a diabetes assistant developed in India and delivered through a widely used messaging application. Expert reviewers rated 86 per cent of its answers as accurate and relevant to context, but the tool scored lowest on neutrality. Dr Dey drew a clear lesson for the field: “Responsible AI starts with auditing your own.” 

Equity ran through the presentation. Many tools are trained on data from Western, higher income populations and may perform less well for women, older adults, ethnic minorities, people in rural and low resource settings, and people who do not speak English. Without deliberate safeguards, Dr Dey warned, AI could deepen rather than narrow existing gaps in diabetes care.

The IDF's position was set out in The Lancet Diabetes and Endocrinology in 2025, in a commentary by Professor Stefan Bornstein, Dr Dey, Dr Charlotte Steenblock, Dr Viswanathan Mohan, Dr Ranjit Mohan Anjana and IDF President Professor Peter Schwarz, titled “Artificial intelligence and diabetes: time for action and caution”. The authors wrote that “these autonomous technological advancements call for responsible adoption of AI in diabetes care, with the clinician in the loop” (Bornstein et al., 2025). To put that principle into practice, the IDF established its Technology and AI Working Group in June 2025 and held regional roundtables in Geneva, Mumbai, Vienna, Singapore
and Dubai between August and December 2025, including a policy session with WHO and other international partners.

Dr Dey closed with a call for shared responsibility. For policymakers, he identified three priorities: supporting the harmonisation of regulation across countries, funding research to close evidence gaps, and protecting vulnerable populations. As he put it, “589 million adults with diabetes deserve AI that is equitable, safe and effective.”

○ ADVOCACY ACTION FOR PARLIAMENTARIANS
Parliamentarians should ask their governments how AI tools used in diabetes care are assessed before they reach patients, and whether they are tested in populations like their own. Members can press for clear rules on who is accountable when an AI tool gives wrong advice, for a requirement that a clinician reviews any AI recommendation on insulin dosing, for systems to report harm linked to AI, and for public funding of independent evaluation. Health and digital committees can also ask how national regulation aligns with international frameworks such as WHO's 2021 guidance on the ethics and governance of AI for health.