Notice: Test mode is enabled. While in test mode no live donations are processed.
Portugal has become the first European Union member state to join the HealthAI Global Regulatory Network, marking an important step towards coordinated international oversight of artificial intelligence in healthcare.
The agreement was signed in Lisbon on 15 July 2026 by Portugal’s National Authority of Medicines and Health Products, INFARMED, and HealthAI—The Global Agency for Responsible AI in Health. The signing took place alongside a major World Health Organization conference on artificial intelligence in health.
According to the [official HealthAI announcement] Portugal’s participation is intended to strengthen cooperation among regulators working to ensure that healthcare AI is safe, effective and equitable. Portugal joins countries including the United Kingdom, India, Brazil, Singapore, Vietnam, Indonesia, Zambia, the Philippines and Peru.
This development is more than a symbolic membership announcement. It offers a practical model for countries attempting to capture the benefits of AI without exposing patients to poorly tested systems, hidden bias, weak accountability or inadequate data protections.
HealthAI is a Geneva-based nonprofit organisation focused on helping countries build the capacity required to regulate artificial intelligence in healthcare.
Its Global Regulatory Network brings national regulators together to exchange knowledge, coordinate approaches and share information about AI-related risks. The network is designed to support governments that might otherwise have to evaluate rapidly changing technologies independently.
Through the agreement, Portuguese regulators are expected to gain access to a directory of regulator-reviewed AI health tools and a real-time warning mechanism through which participating authorities can share reports of harmful incidents. These functions were highlighted in [Reuters’ coverage of Portugal joining the HealthAI network].
A shared directory could help regulators compare evidence, intended uses and known limitations before AI systems are introduced into hospitals or public health programmes. An international warning system could be even more important. When an AI diagnostic, monitoring or decision-support tool produces unsafe results in one country, regulators elsewhere may be alerted before the same problem causes wider harm.
Portugal is not the first European country in the network—the United Kingdom is already a member—but it is the first member from the European Union.
That distinction matters because Portugal will participate while EU institutions and member states continue implementing the European Union’s risk-based regulatory framework for artificial intelligence.
Under the EU AI Act, certain healthcare applications can be treated as high-risk, particularly when AI forms part of a regulated medical device or influences decisions affecting patient safety. The legislation introduces requirements covering areas such as risk management, technical documentation, data governance, human oversight, accuracy, cybersecurity and post-market monitoring.
The implementation schedule has developed over time. The European Commission’s [official AI Act overview] explains that transparency provisions are taking effect in phases, while rules for high-risk AI embedded in regulated products—including certain medical devices—have a longer transition period extending to August 2028.
Portugal’s HealthAI participation therefore gives its regulator an opportunity to compare EU implementation with approaches being developed in Asia, Latin America, Africa and the wider international community.
It may also help reduce regulatory fragmentation. AI developers currently face different evidence, reporting and approval expectations across markets. Greater regulatory cooperation could make those expectations more consistent while still allowing countries to protect local health priorities, patient rights and data sovereignty.
INFARMED regulates and evaluates medicines, medical devices and other health products in Portugal. Its involvement places healthcare AI governance within an institution that already understands clinical evidence, product safety, market surveillance and public health protection.
That is important because healthcare AI cannot be governed solely as ordinary software.
An AI-enabled screening tool, clinical documentation assistant or diagnostic model may change after deployment as software is updated, data sources shift or clinical workflows evolve. Regulators therefore need systems for continuous oversight rather than relying only on a one-time assessment before launch.
The agreement was signed during the WHO’s global conference, “Shaping AI in Health,” co-hosted by WHO/Europe and the Government of Portugal on 15–16 July 2026.
The [WHO conference information] shows that the event brought together governments and healthcare leaders to consider how countries can use AI safely, ethically and effectively. WHO reported participation from 37 countries across all six WHO regions.
Holding the signing during this conference reinforces the idea that AI governance is becoming a global health priority rather than only a technology-policy issue.
Portugal’s decision has implications beyond Europe.
Many health regulators in low- and middle-income countries operate with limited budgets, small technical teams and rapidly expanding responsibilities. They may be expected to assess AI-powered medical devices, chatbots, imaging platforms and digital health applications without having specialised AI evaluation units.
A shared regulatory network can reduce unnecessary duplication. Instead of every authority beginning from zero, countries can exchange evaluation frameworks, evidence requirements, safety alerts and lessons from real-world deployment.
Zambia’s membership is particularly significant because it became the first African country to join the HealthAI Global Regulatory Network. Its participation creates an opportunity for African regulatory priorities to be represented in international discussions, including issues that may receive less attention in high-income markets:
Portugal’s involvement could also create an additional bridge between European institutions and Portuguese-speaking African countries. Such cooperation is not automatic, but it could support future knowledge exchange with countries such as Mozambique, Angola, Cabo Verde, Guinea-Bissau and São Tomé and Príncipe.
Joining an international regulatory network should not be interpreted as approving every AI product available in a participating country.
Regulator-reviewed directories, shared standards and incident-reporting systems are important tools, but they cannot replace local evaluation. A model that performs well in Lisbon may not produce the same results in Maputo, Lusaka or a rural district hospital.
Differences in disease prevalence, equipment, clinical practice, language, connectivity and patient demographics can substantially affect performance.
Countries must therefore ask:
These questions also apply to conversational tools and AI health assistants. APOIO’s article on [the growing role of AI assistants in healthcare access] explains why accessibility gains must be balanced against risks involving inaccurate advice, privacy, escalation and unequal digital access.
Portugal’s move offers several practical lessons for other countries.
Health ministries and regulatory authorities should create national inventories of AI systems already being used in healthcare. They should introduce clear procurement requirements, adverse-incident reporting processes, local validation standards and rules for human oversight.
Countries with limited regulatory resources should explore regional cooperation rather than attempting to build every technical capability independently.
Donors should fund regulatory capacity as seriously as they fund technology pilots. This includes training regulators, supporting independent evaluations, developing public-interest datasets and strengthening ethics committees.
Funding should also cover implementation infrastructure, including connectivity, cybersecurity, clinical workflow redesign and community engagement.
Technology companies should provide evidence that goes beyond promotional accuracy claims. Regulators and purchasers need information about training data, intended use, excluded populations, known failure modes, model updates and post-deployment monitoring.
Developers must also demonstrate that their products work in the settings where they will actually be deployed.
Portugal becoming the first EU state to join the HealthAI network is an important sign that healthcare AI governance is moving towards international cooperation.
The most promising aspect of the agreement is not simply access to more AI tools. It is the creation of mechanisms through which regulators can compare evidence, share safety information and respond collectively to emerging risks.
For Europe, Portugal can help connect the EU’s regulatory framework with wider global experience. For Africa and the Global South, the network could provide access to regulatory knowledge that would otherwise be expensive and time-consuming to develop independently.
However, equitable outcomes will depend on meaningful participation from low- and middle-income countries. Global standards must reflect diverse health systems, patient populations and resource constraints. They must not become rules designed exclusively by wealthy markets and exported elsewhere without adaptation.
Portugal’s decision is therefore a beginning rather than an endpoint. The real measure of success will be whether international cooperation helps governments introduce healthcare AI that is locally validated, affordable, transparent and accountable—and whether those systems ultimately improve access and outcomes for patients who have historically been left behind.
Discover how AI, smart yoga and genomics are making healthcare more personalised, preventive and accessible.
Anthropic’s move into drug discovery could reshape AI healthcare, neglected disease research, and access to medicines.
Robotic drug testing helps scientists detect hidden cancer cells that survive treatment and may cause relapse.
