# Accessed, never disclosed: Silent access to protected health information by medical artificial intelligence agents

DevFeed: [Accessed, never disclosed: Silent access to protected health information by medical artificial intelligence agents](<https://devfeed.tech/articles/accessed-never-disclosed-silent-access-to-protected-health-information-by-medical-artificial-intelligence-agents-114074.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.02319>)

Author: Jae Wang Lee, Abhirup Banerjee, Yeong Chan Lee

Published: 2026-11-03T05:00:00Z

Content type: news

Language: en

Sources: [arXiv: Artificial Intelligence](<https://devfeed.tech/sources/arxiv-artificial-intelligence.md>), [arXiv: Cryptography and Security](<https://devfeed.tech/sources/arxiv-cryptography-and-security.md>)

Topics: [ai security](<https://devfeed.tech/topics/ai-security.md>), [Artificial Intelligence](<https://devfeed.tech/topics/ai.md>), [least privilege](<https://devfeed.tech/topics/least-privilege.md>), [ai evaluation](<https://devfeed.tech/topics/ai-evaluation.md>), [sensitive data](<https://devfeed.tech/topics/sensitive-data.md>)

Tags: [artificial-intelligence](<https://devfeed.tech/tags/artificial-intelligence.md>), [artificial-intelligence-agents](<https://devfeed.tech/tags/artificial-intelligence-agents.md>), [audits](<https://devfeed.tech/tags/audits.md>), [cs-ai](<https://devfeed.tech/tags/cs-ai.md>), [cs-cl](<https://devfeed.tech/tags/cs-cl.md>), [cs-cr](<https://devfeed.tech/tags/cs-cr.md>), [cs-cy](<https://devfeed.tech/tags/cs-cy.md>), [data-privacy](<https://devfeed.tech/tags/data-privacy.md>), [evaluation](<https://devfeed.tech/tags/evaluation.md>), [health](<https://devfeed.tech/tags/health.md>), [medical](<https://devfeed.tech/tags/medical.md>), [privacy](<https://devfeed.tech/tags/privacy.md>)

## AI overview

In tests on identifier-free clinical tasks using synthetic patient records, three of five language models unnecessarily accessed PHI tools, while Claude Opus 4.8 and GPT-5.5 did not. Stronger boundaries reduced unnecessary access; under the strictest policy, remaining access occurred only when users overrode the policy or claimed approval. A pre-execution gate applying a task-necessity criterion nearly eliminated access. Because output-only checks miss access that is not disclosed in answers, the authors conclude that boundaries, gates, and access-log audits are needed together.

## Source excerpt

arXiv:2611.02319v1 Announce Type: cross Abstract: Medical artificial intelligence agents should retrieve patient identifiers only when a task requires them, the data-minimization principle behind the HIPAA minimum necessary standard for protected health information (PHI). An agent can retrieve identifiers without revealing them. Such silent access remains invisible to evaluations that inspect only the output. We embedded canaries in synthetic patient records and linked each tool invocation to canary exposure in the final answers across five language models, four boundary levels, and seven attack frames for identifier-free clinical tasks. Three models invoked the PHI tools unnecessarily, whereas Claude Opus 4.8 and GPT-5.5 did not. Stronger boundaries reduced this access, and the residual access under the strictest policy was confined to frames in which the user overrode the policy or asserted approval. A pre-execution gate given the task-necessity criterion nearly eliminated this access. Output-only evaluation cannot verify minimum-necessary access; boundaries, gates, and access-log audits are required together.