Artificial Intelligence guidelines beyond the medical checklist

In his closing keynote at the recent Asia-Pacific Association of Medical Journal Editors Conference held in Manila, professor Jose Lapeña, president of the World Association of Medical Editors (WAME), rightfully cautioned against treating artificial intelligence (AI) guidelines like a simple checklist, a mechanical tick-box exercise that offers false security while ignoring deeper ethical questions.

He is right, but the reality of safety checklists in health care is telling. As a former nursing student at the Philippine General Hospital, I saw how the surgical safety checklist served as a constant beat in a chaotic operating complex. Yet, surgical complications still happen when steps are skipped under time pressure, and boxes are checked merely to satisfy paperwork. A checklist taped to a chart cannot guarantee safety if the surrounding culture is broken.

We already have our baseline guidelines. Organizations like WAME recommend disclosing AI tools, detailing prompts, and prohibiting chatbots from receiving authorship. The critical question isn’t whether researchers need guidance, but who can actually afford to follow it.

When safety steps are missed in an operating room, it is usually a breakdown of vigilance, because the necessary equipment is already present. In academic research, however, institutions often respond to confidentiality concerns by adopting secure, enterprise-grade AI tools. That solves the problem for well-funded organizations but leaves under-resourced ones behind.

An author at a well-resourced institution gets a compliant, secure workflow. An author at an under-resourced one faces a challenging choice: use an unsecured public model and risk confidentiality breaches, or forego the tool entirely and lose its leveling effect. We cannot demand compliance with a safety checklist when safe tools aren’t provided.

WAME rightly states that detection tools should be accessible to editors “regardless of ability to pay.” That exact logic must be extended one step earlier in the pipeline, to the tools that allow authors to comply in the first place.

We do not need another set of numbered recommendations. We need existing ones to be feasible for everyone, not just those whose institutions can afford them. This is not a technology problem. It is the same funding and infrastructure gap that has long forced local researchers to navigate administrative hurdles and administrative lag. AI did not create this divide, but it is the newest arena where it manifests.

Our researchers continually push forward despite these structural gaps. It is time for our institutions and publishing systems to prove these researchers are worth investing in by building an ecosystem where compliance is not a privilege reserved for the few but a baseline that is supported.

REINER LORENZO TAMAYO,
renztamayo@gmail.com