26th International AIDS Conference 2026 in Rio de Janeiro
Article by Walter Ceron

Artificial intelligence (AI) is increasingly establishing itself as a new tool for the prevention, early detection and treatment of HIV. At a session dedicated to this topic at the International AIDS Conference, several research teams presented specific examples of AI applications, emphasising that it would never replace healthcare professionals. Its role is to improve clinical decision-making, identify people in need of support more quickly and optimise the use of available resources.

One of the most notable projects focused on the prevention of mother-to-child transmission of HIV. Researchers have developed an AI model that can assess the risk of viral transmission to the newborn even before birth. The model takes into account various clinical parameters, including the timing of the start of antiretroviral therapy, viral load, CD4 cell count and adherence to treatment. The researchers have shown that this information not only enables the identification of high-risk pregnancies but also allows for the implementation of targeted measures that can significantly reduce the risk of transmission.

A second presentation focused on the use of AI to identify individuals who have been diagnosed with a sexually transmitted infection (STI) and are at high risk of contracting HIV. The researchers emphasised, however, that the technical quality of the model is only part of the challenge. Acceptance by the communities concerned is equally important. Interviews with LGBTQ+ individuals revealed concerns regarding data protection, confidentiality and the risk of stigmatisation. The participants emphasised that these tools will only fulfil their full potential if they lead to easier access to testing, PrEP or other prevention services.

The session also addressed the use of AI for the automated analysis of chest X-rays in the early detection of tuberculosis in people living with HIV. The speakers pointed out that an algorithm’s performance depends heavily on the population used for its development and that the models must be specifically validated in people living with HIV to avoid misdiagnoses.

Finally, several innovations relating to connected self-tests were presented. Using a photo taken with a smartphone, artificial intelligence can automatically analyse the result of an HIV self-test and immediately refer the person concerned to the appropriate services. Researchers see this as an opportunity to improve early detection and access to healthcare, particularly in contexts where access to healthcare facilities remains limited.

Transparent and responsible AI

Beyond the technical achievements, one message ran through the entire session:

Artificial intelligence must remain a decision-support tool and must not be a substitute for clinical judgement. The speakers emphasised the need to develop transparent models that can explain their predictions whilst ensuring data protection and the involvement of the communities concerned. This approach now appears to be an essential prerequisite for the sustainable integration of AI into HIV prevention and care services.