In a new publication, the World Health Organization (WHO) presents a number of issues that should be taken into account Regulating artificial intelligence for health. The publication emphasizes the importance of establishing the effectiveness and safety of artificial intelligence systems, promptly making these systems available to those who need them, and encouraging dialogue between interested parties, including developers, regulators, manufacturers, healthcare professionals and patients. to promote.
With the increasing availability of health information and rapid advances in analytical techniques (be it machine learning, logic or statistical models), artificial intelligence tools could transform the healthcare sector. WHO recognizes the potential of artificial intelligence to improve health outcomes by strengthening clinical trials, medical diagnosis, treatments, personal care and person-centered care, and by supporting the knowledge, skills and competencies of health professionals. For example, it could be beneficial in settings with a shortage of medical professionals, whether for interpreting retinal scans and radiology images or for other purposes.
However, AI technologies, including large language models, are being deployed rapidly, sometimes without a full understanding of how they work, which could benefit or harm end users, including healthcare professionals and patients. When using health data, these systems could have access to sensitive personal data, requiring strict legal and regulatory frameworks to protect privacy, security and integrity. This is what this publication aims to help promote and sustain.
“Artificial intelligence holds great promise for health, but it also poses significant challenges, including unethical data collection, cybersecurity threats, and the rise of bias and misinformation,” said Dr. Tedros Adhanom Ghebreyesus, Director-General of the WHO. “These new guidelines will help countries effectively regulate artificial intelligence, harness its potential to treat cancer or detect tuberculosis, and minimize risks.”
In response to the growing need for countries to responsibly regulate the rapid development of AI-based health technologies, the publication highlights six areas:
- To build trust, the importance of is emphasized transparency And The documentationwhich means documenting the entire life cycle of the product and recording its development processes.
- refer to Risk managementThere are issues such as “intended use”, “continuous learning”, human intervention, training models and cybersecurity threats that need to be comprehensively improved and simplified as much as possible.
- The external data validation and clarity about it Expected usage Artificial intelligence helps ensure security and facilitate regulation.
- The commitment to Data qualitywhich can be demonstrated through a rigorous assessment of systems prior to their release, is critical to ensuring that systems do not spread bias and misinformation.
- The challenges posed by complex and important regulations such as the General Data Protection Regulation in Europe and the Health Insurance Portability and Accountability Act in the United States of America are addressed, with a focus on understanding jurisdiction and consent requirements. in the service of privacy and data protection.
- Promote this Cooperation between regulators, patients, healthcare professionals, industry representatives and government partners can help ensure products and services remain compliant throughout their lifecycle.
Artificial intelligence systems are complex and depend not only on the code used to create them; The data with which they are trained is also influenced, which comes from, among other things, clinical environments and interactions between users. Improved regulation can help control the risks of artificial intelligence escalating existing biases in training data.
For example, it can be difficult for AI models to accurately represent the diversity of populations, leading to biases, inaccuracies, or even errors. To mitigate risks, regulations can be used to capture characteristics such as gender, race and ethnicity of people in training data and to specifically structure data sets in a representative way.
The new WHO publication aims to set out key principles that governments and regulators can follow to develop new or adapt existing guidelines on artificial intelligence at a national or regional level.