# Athenahealth's Claude link: why HIPAA may not cover your records

> Athenahealth lets patients connect medical records to Claude via MCP. Lawyers say HIPAA may fall away once the data leaves the EHR.

- Published: 2026-11-04
- Topic: [AI](https://futuretechnologyhq.com/topic/ai/)
- URL: https://futuretechnologyhq.com/article/athenahealth-s-claude-link-why-hipaa-may-not-cover-your-records/
- Publisher: [Future Technology HQ](https://futuretechnologyhq.com/), free to read. Quote as "Future Technology HQ" with a link to the article.

## Key points

- HIPAA generally does not follow patient-directed data into a consumer AI service, so the AI provider's privacy policy becomes the main protection
- Athenahealth says patients control what is shared, can revoke access at any time, and that session data is retained only for the session
- Psychotherapy notes and 42 CFR Part 2 records should not go into these systems, according to attorney Rachel Rose
- Healthcare IT teams should map whether an AI arrangement is patient-directed or contracted on behalf of a regulated organisation

## Article

When a patient hands their medical record to an AI chatbot, does the law that normally protects that record still apply? For athenahealth's new Claude connection, the answer from the company and from several lawyers is: generally not, and that gap is the story.

On Monday, the Boston-based electronic health record (EHR) vendor unveiled athenaPatient MCP, which lets patients authorise Anthropic's Claude assistant to pull information from their athenahealth records. That covers laboratory results, medications, diagnoses, allergies, immunisations, medical procedures and care plans.

## What athenaPatient MCP actually is

MCP stands for Model Context Protocol, an open standard for connecting AI applications to external data sources and systems. It is a standard way for AI applications to connect to outside data sources and systems.

Initially the feature is available to US subscribers of Claude Pro and Max. Athenahealth says it is meant to help patients understand their records, prepare questions for doctors and other clinicians, and become more engaged in their care. The timing fits a wider pattern: a Pew Research Center survey cited in the source found that 34% of US adults had used AI chatbots for health-related purposes, and in September OpenAI announced integration capabilities bringing authorised patient information from Epic EHRs into ChatGPT.

## How it works

A patient switches on the athenaPatient MCP connector in Claude's settings and signs in with their athenahealth patient portal credentials. From there, Claude can retrieve that patient's own records and answer natural-language questions about them, such as what a lab result means.

Sam Lambson, athenahealth's data and ecosystem leader, told ISMG that Claude's access is limited to the authenticated patient's records and governed by the same field-level access controls used by the athenaOne platform. The master record stays inside athenahealth's systems. Data is encrypted in transit, then processed and stored within Anthropic's systems for the duration of the session, and Lambson says it is retained only for that session. Patients choose whether to connect, decide what is shared, can revoke access at any time, and can adjust permissions for categories such as sensitive records.

## Why HIPAA may not follow the data

Lambson was blunt: "This is patient-directed, and the AI assistant is not a covered entity or business associate." That does not mean the data is unprotected, he said: "It means that HIPAA is not applicable in this use case."

Regulatory attorney Jordan Cohen of Akerman LLP offered an analogy. The MCP connector is a "digital courier" the patient sends to fetch their records. It resembles downloading records from a portal and uploading them to cloud storage; the storage provider does not become HIPAA-regulated. Federal law requires HIPAA-regulated entities to comply with the transfer request, but afterwards "HIPAA generally falls away", and no business associate agreement is needed.

The picture changes if the arrangement changes. Attorney Rachel Rose noted that athenahealth is itself a HIPAA business associate. If it contracted directly with an AI provider to process patient information on behalf of a regulated organisation, different rules could apply. That distinction could decide whether HIPAA, the Federal Trade Commission Act, the FTC Health Breach Notification Rule or other privacy laws bite. State law matters too: Rose pointed to Texas House Bill 300, whose definition of covered entity is broader than HIPAA's and can reach businesses that collect, store or transmit protected health information.

## What the limitations are

The protection patients are left with is the AI provider's privacy policy plus federal and state consumer privacy law. Rose warns that psychotherapy notes and 42 CFR Part 2 records on substance use disorders should not be put into these systems at all.

Maggie Hales, an attorney and CEO of ET&C Group LLC, raised a separate risk: accuracy. Patients may not appreciate the potential harm from a chatbot's inaccuracies, biases and hallucinations, which she said could lead to claims of negligence and malpractice. Her question has no settled answer: "If patients rely on AI interpretations to their detriment, who will be held responsible?" She argues AI development has outpaced any policy or governance framework.

## The UK angle

The product is initially for US Claude subscribers, and the source gives no UK availability or price. Still, UK readers in health IT should note the principle. Patient-directed sharing sits outside HIPAA, but in the UK, data protection law would be assessed under UK GDPR and the Data Protection Act 2018, with the ICO as regulator. UK health suppliers and NHS-adjacent organisations eyeing similar connectors should treat the question of whose behalf the AI acts on as a design decision, not an afterthought.

## What this means in practice

For patients, Cohen's checklist is the sensible starting point: how long data is kept, whether it trains models, who can access it, and how to revoke access and delete what was shared. Mental health, reproductive health and genetic records deserve extra care. Athenahealth itself advises reviewing the privacy and data-use policies of any third-party service before authorising it.

For healthcare IT and compliance teams, the first job is classification. Is a given AI integration patient-directed, or contracted on behalf of the organisation? The answer decides which laws apply, whether a business associate agreement is needed, and what you must tell patients.

## Where this is heading

The legal position rests on one distinction, patient courier versus provider contractor, and regulators have yet to test it.

## Sources

- [bankinfosecurity.com](https://www.bankinfosecurity.com/athenahealths-claude-ai-connection-raises-hipaa-questions-a-33284)
- [ICO, Guidance on AI and data protection](https://ico.org.uk/for-organisations/uk-gdpr-guidance-and-resources/artificial-intelligence/guidance-on-ai-and-data-protection/)
- [NCSC, AI and cyber security: what you need to know](https://www.ncsc.gov.uk/guidance/ai-and-cyber-security-what-you-need-to-know)
- [govinfosecurity.com](https://www.govinfosecurity.com/athenahealths-claude-ai-connection-raises-hipaa-questions-a-33284)
