Federal health officials convened leaders from 10 technology companies at a private event inside FDA headquarters last month as the agency and CMS work to expand the use of artificial intelligence in clinical care.
STAT reported that a clinical AI demo day took place on July 8 at FDA's White Oak campus, citing an agenda for the event. The companies listed were Anthropic, Counsel Health, Curai, K Health, Microsoft AI, Amazon One Medical, Doctronic, Ellipsis Health, Hippocratic AI and Welldoc. STAT reported that the meeting was intended to give officials firsthand experience with AI doctor technology as it exists today, and that industry, researchers, and lobbyists have been invited to closed-door sessions on the subject.
The question this raises is not whether the meeting was lawful. It plainly was. Agencies consult regulated industry constantly, and doing so is a normal part of understanding a fast-moving technology. The question is disclosure: what the public learns about these conversations, and whether the clinicians and patients who bear the consequences of clinical AI have comparable access to the same officials.
Private Consultation Is Routine, and That Is the Point
FDA meets with sponsors and manufacturers as a matter of course. Pre-submission meetings, roundtables and technical exchanges are built into how medical products are reviewed, and confidentiality often protects legitimate commercial information about unreleased products.
What distinguishes an adoption policy discussion from a product review is the subject matter. A pre-submission meeting concerns one company's device. A convening about how to accelerate clinical AI adoption across the health system concerns policy that will apply to everyone, and policy discussions generally carry stronger expectations of public participation.
The agencies have public mechanisms for that. FDA runs advisory committees, including a Digital Health Advisory Committee, publishes draft guidance for comment, holds public workshops and maintains dockets.
Some broader consultation is underway. STAT has also reported that the White House Office of Science and Technology Policy, FDA and the federal health IT office invited outside experts to help develop a consensus set of principles for benchmarking and evaluating clinical AI. Whether that process reaches the same officials, or carries the same weight, as a demonstration day inside FDA headquarters is not something the public record answers.
Clinicians Have Been Asking for a Seat
Physician organizations have been explicit about what they want from health AI policy, and it is not exclusion from the room.
At its 2026 annual meeting, delegates to the American Medical Association adopted policy establishing that AI should serve as an assistive tool rather than an autonomous decision maker. The policy opposes the use of autonomous or semiautonomous AI systems as substitutes for physician review in coverage determinations and calls for regulations requiring AI-enabled technologies to be integrated into physician-led processes. Delegates also directed the organization to seek regular audits of AI-driven clinical review tools and standards for evidence attribution and transparency in clinical decision support.
"AI has enormous potential in healthcare, but it cannot replace physician judgment," said AMA chief executive Dr. John Whyte. A separate resolution recommended that autonomous clinical AI, including AI prescribing services, be regulated and licensed by an appropriate body, developed with physician input and operated under direct physician supervision.
Those are the positions of the people who will be held responsible when an AI tool contributes to a clinical error, and they are also the people best positioned to identify where a tool fails in practice rather than in validation. Patient advocacy groups occupy a similar position. The harms from a poorly validated clinical AI tool fall on patients first, through missed diagnoses, inappropriate triage, or automation bias in which a clinician defers to a confident but wrong recommendation.
The Policy Backdrop Is Already Moving Toward Less Oversight
This meeting did not occur in a vacuum, and the direction of surrounding policy is what gives the transparency question weight.
Federal policy on AI has turned notably deregulatory. FDA revised its clinical decision support guidance in January 2026 in a way that gives AI-enabled decision-support tools more room to operate, provided clinicians can independently check the reasoning. States have moved the other way, passing more than a dozen laws in 2026 governing health AI, most of which require a human to sign off before an AI-assisted decision reaches a patient. Malpractice law has not settled who is liable when a model's suggestion turns out to be wrong.
When federal oversight requirements loosen at the same time that adoption policy is being shaped in private with the companies that benefit from adoption, the disclosure question becomes more consequential rather than less. That is an observation about process, not an accusation about motive.
The Practical Stake for Patients
For a household, none of this is abstract, and it is worth translating.
AI is already embedded in care in ways patients rarely see: triage tools that route symptom calls, documentation systems that draft clinical notes, imaging algorithms that flag findings, and chatbots that answer questions before a clinician does. Most patients are not told when a tool contributed to a decision about them.
The reasonable things to do are modest. Patients can ask whether an AI system was involved in a recommendation and how a clinician verified it, and can ask for a human review of any automated determination that affects their care or coverage. Anyone using a consumer health chatbot should treat it as information rather than diagnosis, and should not delay evaluation of urgent symptoms because an app was reassuring.
Chest pain, difficulty breathing, signs of stroke, or a rapidly worsening condition require emergency care regardless of what any tool says.
What remains unknown is whether agendas, attendee lists, or summaries from the July event will be released, whether comparable sessions with clinicians and patient groups are planned, and what policy emerged from the discussion. MedicalDaily will monitor FDA and CMS public dockets and advisory committee calendars for follow-through.
Disclosure: Anthropic, one of the companies listed on the event agenda, is named here as a matter of public reporting. MedicalDaily has no financial relationship with any company named in this article.
Frequently Asked Questions
What happened? FDA and CMS officials hosted leaders from 10 technology companies at a clinical AI event at FDA headquarters on July 8, according to an agenda reviewed by STAT.
Was the meeting illegal? No. Agencies routinely consult regulated industry. The issue raised is disclosure, not legality.
Which companies attended? Anthropic, Counsel Health, Curai, K Health, Microsoft AI, Amazon One Medical, Doctronic, Ellipsis Health, Hippocratic AI and Welldoc.
Were clinicians or patient groups included? The reported agenda lists technology companies. A separate federal process has invited outside experts to help set principles for evaluating clinical AI.
Is AI already used in my care? Very likely, in triage, documentation, imaging review, and patient-facing chat tools.
Can I ask about it? Yes. You can ask whether an AI tool contributed to a recommendation and request human review of an automated determination.
Should I rely on a health chatbot? Treat it as information, not a diagnosis. Seek emergency care for severe or rapidly worsening symptoms regardless.
