Doctolib changes privacy policy for AI training with anonymized data

Contents

Doctolib, the company known for its appointment service, which recently acquired the AI-supported answering machine Aaron.ai, now wants to use patient data to train AI models. To this end, the company has amended its privacy policy, as Netzpolitik.org has discovered.

According to this, in addition to search data, appointment history, documents, medical notes and medical information entered by the user on the platform, voice recordings will also be available for the training of AI models. However, the user's express consent is required for the use of personal health data, as stated in the privacy policy. If "no health data is involved", Doctolib cites "legitimate interest" as the legal basis for data processing.

More and more companies are planning AI training with patient data; Medatixx even offers doctors a discount if they provide data from the practice management system. Big tech such as OpenAI, Microsoft, Meta and Google are also interested in health data to train their AI models.

Training with anonymized data

For its AI projects, Doctolib relies on "proven technologies" such as Whisper (speech-to-text-to-speech) and GPT-4o (text-to-summary) from OpenAI in addition to AI models developed in-house, according to the company when asked by heise online. AI models would only be "trained and further developed with anonymized data after explicit consent". The company also promises to strictly adhere to data protection guidelines and only store data on servers in Europe.

Patient file and more planned

Doctolib now has over 80 million users across Europe and has announced the launch of an electronic patient file for France, which is seen as a potential competitor to the state version, "Mon espace santé". This was reported by the Huffington Post at the end of 2024. Doctolib, which originated in France, is also widely used in Germany. Doctors' surgeries often only offer appointments via Doctolib.

On the occasion of European Data Protection Day, Hessian Data Protection Commissioner Alexander Roßnagel also appealed for consideration for all those "who cannot or do not want to go along". According to Roßnagel, the use of an app that forces users to disclose personal information is not compatible with the General Data Protection Regulation. He is not in favor of making appointments via an AI chatbot.

Long waiting times

The demand for services such as Doctolib, which already has a history of data protection issues, is fueled by difficulties in allocating appointments –, especially for doctors in private practice. There are various reasons for this, such as staff shortages in doctors' surgeries. The National Association of Statutory Health Insurance Funds recently published a survey: 42 percent of those surveyed with statutory health insurance stated that waiting times for doctor's appointments had worsened over the past five years. 44 percent of those insured are also dissatisfied with the availability of doctors' surgeries by telephone. 3512 people with statutory health insurance aged between 18 and 80 were surveyed via an online panel between 05.03.2024 and 29.04.2024.

"To improve appointment allocation, we are proposing a statutory regulation that would require all medical practices to make a fixed proportion of their SHI appointments available on a daily basis on an online portal," says Stefanie Stoff-Ahnis, Deputy Chairwoman of the Executive Board of the National Association of Statutory Health Insurance Funds. Practices should also no longer ask whether someone has private or statutory health insurance when making appointments.

The spokesperson for the National Association of Statutory Health Insurance Physicians explained to the Ärzteblatt that with one billion doctor-patient contacts per year, there is "low-threshold access to medical care" in Germany. A large proportion of these are said to be treatments in doctors' surgeries.

(mack)