The AMA just said AI vendors share the blame when the AI is wrong
Plus athenahealth's Claude link for patients, and the FDA's next move on generative AI.
 | No. 003 · Nov 4, 2026 |
01The big story
Who pays when an AI tool gets it wrong
| 3 | parties the AMA says can share liability for clinical AI: the doctor, the company that built the tool and the organization that put it in place. AMA framework, Oct. 29, 2026 |
Doctors are being asked to act as the safety net for clinical AI. They may be held accountable for risks they cannot control, Healthcare Dive reports.
The American Medical Association wants that changed. Its new framework says several parties can share liability, each one accountable for the harm it was in a position to prevent.
That brings in the company that built the tool and the organization that put it in place. Those organizations include health systems, hospitals and physician practices, so your own group can be one of them.
In the AMA's view, the organization that puts a tool in place should test it before asking doctors to use it, and train them on it.
The AMA's reason is that current liability rules were not designed for tools where doctors, developers and health organizations each control a different part.
The framework is a call for change, not a law. Until the rules catch up, your own malpractice policy is the answer that counts.
For a doctor who rounds at more than one hospital, the question comes up at every one. Section 03 shows what to ask your carrier.
02New tools
athenahealth now lets patients connect their chart to Claude
| What | athenaPatient MCP is a connector: it lets a patient link their official health record to Claude, Anthropic's AI chatbot, so Claude answers their questions from the real chart instead of guessing. |
| Evidence | It is built on MCP, the Model Context Protocol, an open standard for linking AI tools to outside data, the company says (HIT Consultant). No independent study is cited. The article covers privacy, with no data kept or used for training, but not how accuracy is checked. |
| Price | athenahealth lists none. It works only for U.S. patients on Claude Pro or Claude Max, which are paid plans. |
| For | Patients, not doctors. If your practice runs on athenahealth, expect some patients to arrive with Claude's reading of their own chart. |
03Try this this week
Nothing below uses patient information.
1. Ask your malpractice carrier who pays when AI is wrong. Ask in writing whether your policy covers AI tools at all. Then ask who it names as responsible when one of them is wrong.
2. Ask for the testing behind any AI tool added to your hospital's chart. The AMA says the hospital or practice that rolls out a tool should test it before doctors use it. Ask your medical staff office for the validation data on any AI tool rolled out this year, not just the marketing sheet.
3. Have a one-line disclosure ready for patients. In a new Pew Research Center survey of U.S. adults, people were open to AI handling tasks like scheduling and visit notes, but 4 in 10 said AI should not make diagnoses, even with a person reviewing it (Healthcare Dive). Use any AI tool to draft wording you can say out loud:
Write one sentence a physician could say to a patient to disclose that an AI tool helped draft a clinical note, in plain language, without naming any specific product.
04Quick hits
The FDA and CMS are looking to work with states on how health AI is regulated (STAT).
The FDA is weighing how to regulate generative AI in medical devices after hundreds of public comments from patients, providers and medtech companies (Healthcare Dive).
Private equity firms now own nearly 1 in 8 private U.S. hospitals, according to a tracker from the Private Equity Stakeholder Project (HIT Consultant).
Is this useful? Hit reply and tell me what you'd want covered next.
Staff is free, and presented by BLKBAG, healthcare software for doctors.
Don't miss what's next. Subscribe to Staff: