Public health gets a protocol for deciding when (not) to use AI in its communications

By Zendoric · October 10, 2026.

The Public Health Communications Collaborative (PHCC), in collaboration with Health Resources in Action (HRIA), published on October 9 a downloadable document called "AI with Intention: The TRAPP Framework". According to the organization itself, the aim is to help public health communications professionals assess "when and how" AI tools should be used in specific tasks in their work, without the source specifying which tasks those are.

The material available on the page is, it must be said, minimal: a publication notice and a download link, with the text itself not spelling out what each letter of TRAPP stands for or what specific criteria it proposes. It is not a piece of research or a study with original data, but a training resource aimed at professionals in the sector, of the kind the PHCC Academy has been producing to strengthen the communication capacity of US public health.

Even with that sparseness, the fact itself deserves context. Public health communication is one of the areas where the margin for error with generative AI is narrowest: a poorly calibrated message about a vaccine, an outbreak or a medicine not only misinforms, it can feed distrust in an institution already dealing with skepticism accumulated since the pandemic. In that terrain, the relevant question is not whether AI can draft a press release -it can-, but whether it should do so without human oversight at every step, from fact-checking to the tone used toward vulnerable communities.

Our reading is that frameworks of this kind -narrow, practical, designed for the working professional rather than the regulator- are how AI governance is actually being built in sensitive sectors: not through a federal law dictating which model to use, but through internal protocols, criteria guides and professional training that arrive ahead of the rule. It is consistent with something we have been observing in other areas: the effective brake on problematic uses of AI tends to arrive through channels other than AI-specific legislation, whether general law applied by a judge or, as here, professional standards that an organization drives within its own field.

In the short term, initiatives like this are necessary precisely because they acknowledge the problem head-on: AI in health communication can save time for chronically under-resourced public health teams, but used badly it erodes the very trust those teams need for people to get vaccinated, attend screenings or follow a health alert. In the long term, if these frameworks mature and are refined, well-governed AI can multiply the reach of health communication -translating notices into dozens of languages instantly, tailoring the message to each community, freeing public health professionals from administrative work so they can devote more time to direct contact- in line with the abundance we advocate for the sector as a whole. But that horizon is earned with frameworks like TRAPP, not taken for granted.