WHO launches guide for ethical use of artificial intelligence in health

The World Health Organization (WHO) today launched a guide with new ethical guidelines to promote and guarantee safe use in the healthcare field of large language models (LMM), a type of artificial intelligence (AI) generative in rapid development.

Capable of generating texts and other predictive content from massive data processing, LMMs, like the controversial tool Chat-GPTadvance with a presence in our lives exponential to the advances of artificial intelligence.

This, recalls the WHO, makes its use increasingly possible in medical settings such as diagnosis, investigation of symptoms and treatments, drug development or health administration tasks.

Although LMMs are beginning to be used for specific health purposes, the WHO warns that some of these utilities are not yet tested nor is it known whether they will offer the expected benefits, which makes it necessary to establish certain standards for their integration and use for medical and of public health.

For this reason, the guide documents some of the potential risks for health systems, such as the so-called “automation bias”which could produce false, inaccurate, biased or incomplete statements, harming people who use said information to make health decisions.

Vulnerable to cyber attacks

LMMs, like other forms of AI, are also vulnerable to cybersecurity risks that could jeopardize patient information or the reliability of these algorithms and the provision of medical care, reminds the health organization of the UN.

Given this, the first step proposed by the WHO for a safe and effective application of LMMs in the health sector is a participation “vertical and integrated” from all stakeholders, including governments, technology companies, healthcare providers and civil society.

At the government level, investment in public infrastructure accessible to AI developers is recommended in exchange for adhering to established ethical principles.

The WHO also advises governments to enact laws to ensure that LMMs comply with human rights, or to assign a regulatory agency to assess whether these technologies meet requirements before use.

In this sense, the director of the WHO Digital Health and Innovation department, Alain Labrique, assured in the presentation of the guide that “Governments of all countries must cooperatively lead efforts to effectively regulate the development and use of AI technologies, such as LMMs.”

According to the WHO, efforts should also be increased in all phases of development of large language models towards a design “structured, inclusive and transparent.”

Participation of the entire society

Not only engineers and scientists, but also potential users and all direct and indirect stakeholders, including medical service providers, health professionals and patients, must participate.

Developers of these technologies will also need to be able to predict and understand the possible side effects of AIs and design them to perform well-defined tasks with the precision and reliability necessary to improve the capacity of healthcare systems and promote the interests of patients.

However, despite the current challenges that LMMs represent for the health sector, the WHO also recognizes the enormous benefits that LMMs AI could provide health systems, including improving public health and achieving universal health coverage.