Listen to this story.

RCVS HQ

RCVS photo

Britain's Royal College of Veterinary Surgeons is helping to formulate guidelines for the responsible use of artificial intelligence tools in veterinary medicine.

Developers of artificial intelligence tools for veterinarians should disclose whether veterinary professionals were involved in their products' development and have ownership in their businesses, according to guidelines being finalized in the United Kingdom.

The recommendation is one of 23 principles in a draft working paper drawn up by the Veterinary AI Transparency Alliance, a group led by the Royal College of Veterinary Surgeons, the U.K.'s professional regulator, and Digital Practice, a company that provides client communication, scheduling and payment services.

The guardrails are being formulated as regulators and professional associations worldwide attempt to place checks and balances on veterinary AI applications that are fast gaining in popularity and come with risks, such as incorrect diagnoses or ethics questions over the use of client data.

The British draft guidelines were developed over 18 months by a group of veterinary professionals, AI product developers, regulators, insurers, educators and animal owners, the RCVS said in a press release. The draft is open for public comment through 5 p.m. U.K. time Monday.

In the United States, a technology task force of the American Veterinary Medical Association has, for the past nine months, been developing AI best-practice guidelines for veterinarians. Michael San Filippo, an AVMA spokesperson, confirmed the guidelines are still under development. The topic of AI, he added, will be discussed by the AVMA House of Delegates on July 9 at the association's annual convention.

AI applications in veterinary medicine are wide-ranging and include tools that help veterinarians take notes, triage patients, perform diagnostic work and monitor patients' vital signs.

They also can make mistakes. As a case in point, recent research published in the Journal of the American Veterinary Medical Association found that six AI tools that read veterinary radiographs sometimes failed to identify existing health conditions or identified conditions that didn't exist.

The stakeholders who collaborated on the U.K. guidelines agreed that AI tools could enhance veterinarians' capabilities and improve animal welfare, RCVS CEO Lizzie Locket said in the press release. "However, this was balanced by the clear desire to use AI tools with requisite caution, taking into account potential risks, the need to be open and transparent about their use, and the need for professional judgement both when choosing an AI service and when their outputs are used," Locket said.

All 23 of the principles in the guidelines center on information that AI providers should disclose to veterinary professionals. One principle states that developers should say whether any veterinarians, registered veterinary nurses or other relevant subject matter experts were involved in the design, testing, validation or review of their products.

The same principle calls for providers to disclose who owns or controls the tool, including any parent company or commercial entity whose products or services the tool might recommend. That would enable users to judge whether the tool's outputs might be influenced by commercial interests. "The emphasis is on present accountability rather than a historical list of past contributors," the guidelines add.

Developers are called upon to clearly describe what their tools can and cannot do and the extent to which human oversight is required, including which tasks should remain under veterinarians' control.

Product providers also should alert users to circumstances in which client consent is likely to be required for their tools' use, disclose where any personal client data is stored, who owns that data and whether the data is used for product improvement or uses beyond the immediate service being provided, the guidelines state.

In the U.S., the AVMA's technology and innovation task force met in September to begin identifying and prioritizing the resources it plans to develop for veterinary teams, the AVMA said in November.

The resources will include information on AI literacy, professional accountability, privacy and data stewardship, transparency and consent, validation of tools for veterinary use, documentation, team training and regulatory compliance, San Filippo said. The task force, he added, also is developing "practice-level governance tools, decision-making checklists and client communication resources that will be made available through a web-based resource center and educational programming."

Guidelines are starting to appear in other countries also. In Canada, for instance, the Alberta Veterinary Medical Association in January published professional standards for the use of AI in veterinary medicine. Among five standards listed in that document is that veterinarians remain responsible and accountable for their conduct when using AI tools and that veterinarians should validate and monitor AI performance — including by staying updated on AI advances.