Artificial Intelligence (AI) is no longer a future concept in the pharmaceutical industry. From clinical trial design and medical literature review to adverse event processing and signal detection, AI is increasingly being integrated into medicine development and lifecycle management.
Recognising this rapid adoption, the European Medicines Agency (EMA) and the U.S. Food and Drug Administration (FDA) recently published joint guiding principles on the use of AI in medicine development. While the document does not introduce new regulations, it provides a clear indication of how regulators expect organisations to develop, validate, govern, and monitor AI systems moving forward.
For pharmacovigilance (PV) teams, these principles are particularly important. AI is already being used to automate case intake, support signal detection, identify safety trends, and improve operational efficiency. However, the regulators’ message is clear: AI can support decision-making, but accountability for patient safety remains firmly with the organisation.
Understanding what these principles mean in practice is essential for PV leaders, Qualified Persons for Pharmacovigilance (QPPVs), safety scientists, compliance teams, and technology stakeholders.
Why the EMA and FDA Released Joint AI Principles
The pharmaceutical industry is experiencing an unprecedented increase in data volume.
Pharmacovigilance departments process information from:
- Individual Case Safety Reports (ICSRs)
- Clinical trial data
- Medical literature
- Social media monitoring
- Patient support programmes
- Electronic health records
- Real-world evidence databases
Traditional manual processes are often unable to keep pace with this growing data landscape.
AI technologies offer opportunities to:
- Improve efficiency
- Accelerate data processing
- Reduce repetitive manual work
- Identify patterns within large datasets
- Support earlier detection of safety concerns
However, regulators recognise that AI systems can also introduce risks, including bias, lack of transparency, inconsistent outputs, and challenges related to validation and oversight.
The joint principles aim to ensure that innovation does not compromise patient safety, data integrity, or regulatory compliance.
The Key Message: Human Oversight Remains Essential
One of the most important takeaways for PV teams is that AI is not a replacement for human expertise.
Many organisations are exploring the use of Generative AI, machine learning models, and automation tools to streamline pharmacovigilance operations. While these technologies can significantly improve efficiency, regulators expect appropriately qualified personnel to maintain oversight of AI-generated outputs.
This means:
- Safety assessments cannot be fully delegated to AI.
- Medical review remains a human responsibility.
- Signal evaluation requires expert judgement.
- Benefit-risk assessments must remain scientifically justified.
- Regulatory decisions require accountable oversight.
In simple terms, AI may assist with identifying potential risks, but organisations remain responsible for evaluating and acting on those risks appropriately.
Data Quality Will Determine AI Success
One of the strongest themes throughout the joint principles is data quality.
AI systems are only as reliable as the information used to train and operate them.
For pharmacovigilance teams, poor-quality data can create significant risks, including:
- Missed safety signals
- Incorrect case classification
- Duplicate case processing
- Biased outputs
- Inaccurate trend analysis
Many organisations focus heavily on selecting AI platforms while paying less attention to the quality of their underlying safety databases.
Regulators are likely to scrutinise:
- Data completeness
- Data consistency
- Source reliability
- Data governance controls
- Ongoing quality monitoring
This means that strong data management practices may become just as important as the AI technology itself.
AI Validation Is Becoming a Compliance Priority
Validation has always been a fundamental requirement within regulated pharmaceutical environments.
The introduction of AI does not remove this expectation. In fact, it increases the need for robust validation approaches.
Unlike traditional software, AI systems can evolve over time, particularly when machine learning models continue learning from new data.
This creates unique challenges.
PV teams should be prepared to demonstrate:
- How AI systems were tested
- What datasets were used for validation
- Performance thresholds and acceptance criteria
- Risk assessments performed before deployment
- Ongoing performance monitoring activities
Regulators are increasingly interested in understanding whether organisations can consistently demonstrate that AI systems perform as intended.
Validation is no longer just an IT responsibility; it is becoming a critical pharmacovigilance governance requirement.
Signal Detection Could Change Significantly
Signal detection is one of the areas where AI has the potential to deliver substantial value.
Traditional signal detection approaches often rely on predefined statistical methods and manual review processes.
AI can support signal management by:
- Analysing large datasets rapidly
- Identifying hidden correlations
- Detecting emerging safety patterns
- Prioritising potential risks for review
- Reducing manual workload
However, regulators are unlikely to accept AI-generated signals without clear justification.
PV teams should be able to explain:
- Why a signal was identified
- Which data contributed to the finding
- How the algorithm reached its conclusion
- What review process was applied
This highlights the growing importance of explainability within AI-enabled pharmacovigilance systems.
Explainability Is Becoming a Regulatory Expectation
One of the greatest concerns surrounding advanced AI models is the so-called “black box” problem.
In some cases, AI systems generate outputs without providing a clear explanation of how conclusions were reached.
For pharmacovigilance activities, this creates obvious challenges.
If an inspector asks why a safety signal was escalated or why a case was classified in a particular way, organisations must be able to provide evidence supporting the decision.
Regulators increasingly expect AI-supported processes to be:
- Transparent
- Documented
- Reproducible
- Scientifically defensible
The more explainable an AI system is, the easier it becomes to demonstrate compliance during inspections and audits.
Third-Party AI Vendors Create Additional Responsibilities
Many pharmaceutical companies are not building AI systems internally.
Instead, they are adopting commercial AI solutions offered by software providers and technology vendors.
The joint principles reinforce an important point: responsibility cannot be outsourced.
Even when using third-party solutions, organisations remain accountable for:
- Vendor qualification
- System validation
- Performance monitoring
- Risk assessment
- Regulatory compliance
This means pharmacovigilance teams may need to work more closely with procurement, quality assurance, information technology, and compliance departments when evaluating AI-enabled platforms.
A vendor’s claims alone are unlikely to satisfy regulatory expectations.
Inspection Readiness in an AI-Driven Environment
As AI adoption increases, inspection readiness requirements are likely to evolve.
Inspectors may begin asking questions such as:
- Which pharmacovigilance activities involve AI?
- How was the system validated?
- What controls exist to prevent incorrect outputs?
- How is performance monitored?
- Who reviews AI-generated recommendations?
- What actions are taken when errors are identified?
Organisations that cannot clearly answer these questions may face increased scrutiny.
Developing AI governance frameworks today can help avoid future compliance challenges.
What PV Teams Should Do Now
Rather than waiting for additional guidance, pharmacovigilance teams can take several proactive steps.
Map Current AI Usage
Many organisations are already using AI without formally categorising it as such.
Conducting an inventory of existing AI-enabled tools can provide valuable visibility.
Strengthen Governance Structures
Establish clear responsibilities for:
- AI oversight
- Validation
- Performance monitoring
- Risk management
- Change control
Review Vendor Management Processes
Ensure AI vendors are assessed with the same level of scrutiny applied to other GxP-relevant systems.
Invest in Training
PV professionals do not need to become data scientists, but they should understand how AI systems influence pharmacovigilance activities and where potential risks exist.
Focus on Explainability
Prioritise solutions that provide transparency and support defensible decision-making processes.
The Future of AI in Pharmacovigilance
The EMA and FDA’s joint principles do not signal resistance to AI adoption. In fact, they acknowledge the significant potential of AI to improve medicine development and patient safety.
However, the principles make it clear that innovation must be balanced with governance, transparency, validation, and accountability.
For pharmacovigilance teams, the future is unlikely to involve fully autonomous safety systems. Instead, it will involve intelligent technologies working alongside experienced safety professionals to improve efficiency while maintaining scientific and regulatory rigour.
Organisations that establish strong governance frameworks now will be better positioned to adopt AI confidently, meet evolving regulatory expectations, and maximise the benefits of these technologies without compromising patient safety.
Preparing PV Teams for the Next Phase of AI Adoption
As regulatory expectations around AI continue to mature, pharmacovigilance teams will need to balance innovation with compliance. Success will depend not only on implementing advanced technologies but also on establishing robust governance, validation, oversight, and quality management processes.
Key areas organisations should focus on include:
- AI governance and accountability
- Data quality and integrity
- Validation and performance monitoring
- Explainability and transparency
- Vendor qualification and oversight
- Inspection readiness
Quality & Vigilance Ltd. (Q&V) supports pharmaceutical and healthcare organisations with expertise in pharmacovigilance, quality assurance, regulatory affairs, inspection readiness, and compliance management. As AI becomes increasingly integrated into PV operations, organisations require practical strategies that ensure innovation remains aligned with regulatory expectations while protecting patient safety and maintaining compliance.