I am not generally keen on scaremongering. Nevertheless, it is increasingly apparent that artificial intelligence is entering clinical practice faster than the professional guidance, organisational governance and legal frameworks can keep pace, and we need to act now.
In human healthcare, we are now seeing the consequences through a rise in complaints and legal cases related to AI use. In both human and veterinary clinics, AI systems are influencing diagnoses, treatment recommendations and medical records, yet clinicians often have limited information about how those systems were developed, validated or tested. When something goes wrong, responsibility typically falls on the professional who used the tool.
The warning is stark - using an opaque or inadequately validated AI tool does not remove the veterinary professional’s responsibility for the resulting outcome.
AI complaints and litigation in healthcare
The risk is no longer theoretical. In July 2026, the Health Service Journal reported that professional regulators had received the first complaints concerning alleged inappropriate use of AI by clinicians. Patients are becoming more aware of how AI may be used in their care and more willing to challenge its use.
Complaints may examine questions such as:
- Was the patient told that AI was being used?
- Were they aware how confidential information was processed?
- Were appropriate data security and privacy measures in place?
- Did the clinician verify an AI-generated record?
- Was the system used for its intended clinical purpose?
- Could the clinician understand its outputs, limitations and challenge its recommendations?
These are not purely technical questions. They concern consent, confidentiality, AI literacy, communication, professional judgement, record keeping, procurement, and individual accountability.
However, accountability may not lie solely with the individual. Organisations responsible for procurement and deployment should be able to demonstrate that a technology has been appropriately evaluated. A recently filed US lawsuit provides a clear warning. A former Mayo Clinic director of research operations and AI compliance lead has alleged that she faced retaliation after raising concerns about the governance of AI-related research and technology. In a rush to include AI into operations, concerns were raised regarding patient care, privacy and data manipulation. While the case is ongoing, it illustrates that records, decisions and governance processes that may be examined when the safety of clinical AI is challenged.
Who is liable when medical AI goes wrong?
There is a widening gap between the use of AI and legal frameworks governing responsibility when harm occurs. Where a regulated professional lies between the AI and the instance of harm, they become the bridge in this liability gap.
According to the Medical Protection Society, existing legislation under the UK Consumer Protection Act 1987, which holds producers liable when a defective product causes injury, does not clearly include software and AI systems in its definition of product. This uncertainty may make it difficult for patients to pursue developers, manufacturers or suppliers when an AI system contributes to harm. The clinician or healthcare organisation may consequently become the target for a clinical negligence claim.
This creates a difficult position for clinical professionals as the human-in-the-loop:
Follow an unsafe AI recommendation and the clinician may be criticised for failing to exercise appropriate judgement. Reject an appropriate recommendation and they may later be asked why decision support was disregarded. Refuse to use AI and face criticism where the system typically outperforms humans in diagnostic or treatment accuracy. In all scenarios, clinicians may have had little control over the system’s design and limited access to information explaining how its output was produced.
Describing a clinician as being “in the loop” is therefore not enough. Meaningful human oversight requires the knowledge, evidence, transparency and processes needed to identify when the technology may be wrong.
While there are moves to include AI within product liability frameworks, consultation and legislative reform take time. Any change is unlikely to offer immediate protection to professionals already using AI in clinical practice. Veterinary professionals and businesses should take note - future legal reform cannot substitute for effective AI governance today.
What does AI litigation risk mean for veterinary medicine?
Veterinary AI operates within a different professional, regulatory and legal environment from human medical AI. However, many of the potential routes to complaints and claims are directly transferable.
A veterinary professional could face scrutiny after:
- relying on an incorrect diagnostic or treatment recommendation;
- failing to detect fabricated or inaccurate AI-generated information;
- using a system outside its intended species, population or clinical setting;
- accepting an AI-generated clinical record without adequate verification;
- sharing identifiable client, patient or business data;
- failing to obtain informed consent;
- failing to communicate relevant limitations when AI influenced advice;
- using a tool without understanding its performance and limitations.
I often find a self-driving car analogy helpful when thinking about AI systems. Imagine you get into a self-driving car with another individual - maybe you're stuck at an airport and it's the only car available, so you have no practical alternative choice. The car gives you no safety data or information. It doesn’t inform you that your conversation is recorded and your data is passed to third parties. But you trust that its very presence in the taxi rank means that someone, somewhere has performed due diligence checks to check that it’s safe and secure. What if there is then an accident that involves harm - as a regulated professional should you be solely liable because you did not understand the risks? Or for not explaining the risk to your co-passenger? For failing to understand and articulate the lack of human-override controls? For not scrutinising every decision the autonomous vehicle made? Or should the manufacturer or local authorities be responsible for inadequate testing and safety - including not providing you and your passenger with important information to make a risk-based judgement?
Veterinary practices, large groups and animal health organisations may indeed face risk through inappropriate procurement processes, inadequate staff training, weak data governance, undocumented implementation, or a failure to monitor AI performance after deployment. When an adverse outcome occurs, organisations may need to demonstrate not simply that an AI system was available, but why its adoption was reasonable and what safeguards surrounded its use. Even if there is not currently an established litigation risk, the danger to reputation, staff retention and revenue is reason enough to act now.
Transparency - the key to unlocking safe veterinary AI adoption
It is impossible to manage risk in the absence of clear understanding. However, transparency is a balance. It does not mean that an AI developer must publish its source code or relinquish commercially sensitive intellectual property. It means they must provide enough meaningful information for users and purchasing organisations to make an informed, defensible assessment of the technology.
Depending on the application, this should include:
- the intended purpose and intended users;
- the species, populations and clinical settings evaluated;
- the provenance and relevance of training and validation data;
- clinically meaningful performance measures;
- known limitations and common failure modes;
- requirements for professional oversight;
- data protection and cybersecurity safeguards;
- version control and update processes;
- arrangements for incident reporting and ongoing performance monitoring.
Without this information, a veterinary professional cannot reliably judge whether an AI tool is suitable for a particular animal or clinical problem. A practice cannot design proportionate training or governance. A purchaser cannot compare competing products on the quality of their evidence. And yet, in a 2026 study of 71 veterinary AI products, the mean transparency score was a pitiful 6.4%.
Opacity may protect a company from scrutiny in the short term, but it transfers uncertainty and risk to users and wider veterinary organisations, and - most importantly - may result in harm to patients and clients.
How independent AI validation can reduce risk
AI has substantial potential to improve animal health, support overstretched veterinary teams and broaden access to care. The answer is not to prevent adoption. It is to make adoption safer, more transparent and more defensible.
Vet Validaite is bringing this into practice through independent expert validation of AI technologies developed for animal health applications.
By assessing technologies against recognised best-practice standards, Vet Validaite examines the evidence, intended use, transparency, governance and clinical relevance underpinning AI products for animal health. This gives veterinary professionals and purchasing organisations the clarity needed to safely and confidently integrate AI into their systems and clinical decision making.
Independent validation also provides responsible developers with the opportunity to show their commitment to best practice and distinguish evidence-led products from technologies supported primarily by marketing claims.
No validation process can remove every clinical or legal risk. It can, however, replace uncertainty and opacity with structured scrutiny and provide a stronger basis for procurement, governance and professional decision-making that supports thriving teams and excellence in patient care.