AI scribes raise privacy and consent worries in doctor's offices. During a recent visit to my physician, I was approached about allowing an artificial intelligence assistant to accompany the consultation. The AI was known as Heidi, part of a rising trend of digital scribes being deployed in healthcare settings. These tools function by capturing audio from the interaction, generating a detailed transcript, and organizing it into structured clinical notes for the physician to review. On the surface, this seems like an ideal application for AI. Medical professionals are often burdened with paperwork, which could be efficiently managed by a machine, allowing them to direct more attention to their patients. This arrangement could potentially enhance the quality of care by reducing administrative strain and improving focus.
Despite the appeal of the idea, I chose to refuse. My doctor mentioned he had used Heidi for more than a year and claimed to have a thorough understanding of the technology. He also shared that he had worked with blockchains in the past. However, when I pressed for specifics—such as where the conversation would be processed, what data would be retained, which companies supported the infrastructure, and who owned the service—he struggled to provide clear answers. What concerned me most was that the person asking for my consent could not fully explain the system. My reservations were not about the AI itself; I use AI regularly in my daily life. I was unsettled by the lack of transparency about how my data might be handled and who had access to it.
What happens to the data no one is telling patients about
Last week, Digital Rights Watch launched a campaign advocating for patients to have the legal right to refuse AI scribes without facing denial of medical care. The organization reports that patients are already being turned away for this very reason, a point I can confirm from my own experience. My issue was not with the presence of the AI in the room. My concern centered on the fact that I was being asked to approve a system whose workings were not clearly explained to me. Without understanding where the data was sent, how it was managed, or which other entities might be involved, it felt like I was being asked to sign off on something I could not fully comprehend.
Heidi's funding origins trace back to global venture capital firms and institutional investors, with the Australian superannuation fund HESTA among those listed as contributors. When I raised these concerns, the doctor could not offer much insight. Instead, he insisted that nothing was being recorded, a claim that is technically impossible. For an AI scribe to generate a transcript, the conversation must be captured through a microphone, converted into digital data, and then processed into text. This means the key distinction isn't between 'recorded' and 'not recorded,' but between data that is processed and that which is retained.
The fine print behind convenience in healthcare
Tom Sulston, Head of Policy at Digital Rights Watch, explained that if the original audio is deleted after transcription, it becomes impossible to verify the accuracy of the resulting clinical notes. Artificial intelligence systems are prone to errors—they can mishear words, omit important details, or even fabricate information. While deleting the recording may protect a patient's privacy, it also opens the door to potential inaccuracies in their medical records. Sulston pointed out that this represents a difficult trade-off between two potential risks. Patients are left to choose between privacy and the ability to verify their medical documentation, both of which are crucial to maintaining trust in the system.
Sulston also emphasized the issue of consent. When refusing an AI scribe means losing access to care, the concept of true consent is undermined. For consent to be valid, it must be informed, voluntary, and free from coercion. However, in a clinical setting, patients may be dealing with illness, anxiety, or physical pain, making it difficult to engage with complex privacy policies or data processing details. What is usually offered instead is a simplified version of the process—'It just writes my notes.' This approach may work in the moment, but it fails to address the larger ethical concerns about transparency and control that patients deserve.
The use of AI scribes in healthcare reflects a broader pattern in technology. Consumers often accept convenience at the cost of control—streaming services offer easy access but erode ownership, smart devices automate tasks but depend on distant servers, and AI systems generate decisions without always revealing how they are reached. These tools begin as gadgets but eventually integrate into everyday life, raising critical questions about our dependencies. The real issue isn't whether the technology functions well, but what we're willing to trade in return for it.
In response to these concerns, I have gradually shifted toward self-hosting solutions. My journey began with the desire to reclaim control over my music collection, which had become dependent on subscription services. By storing my music locally, I was able to reduce reliance on third-party platforms. This decision eventually expanded to include my documents, home automation systems, and even some AI-driven functions. The goal was to regain control over my data and reduce my dependence on external services that I couldn't fully trust or understand.
