Russia has seen an eightfold increase in requests to AI for interpreting medical test results. According to the SberHealth service, between May and October 2025, the number of such inquiries grew more than eight times, and the share of users asking the neural network “what to do next” jumped from 4 percent to 41 percent. A similar trend is observed in most other countries.
Against this backdrop, cybersecurity expert Alexey Gorelkin urged Russians not to share confidential personal information — especially medical data — with AI systems.
Why AI Is Not a Doctor and Never Will Be
Gorelkin explains that a neural network can provide only general health advice and lacks the qualifications of a medical specialist. Medicine is highly individual: the same indicator (for example, elevated blood sugar) can suggest very different diagnoses for different people, often requiring opposite actions.
“AI can be a first step for getting general information, but it should never replace a consultation with a qualified doctor,” the expert emphasized.
The Main Risk: Your Data Becomes Training Material
Even if a particular chat is not directly linked to your identity, nearly all major AI services (including those in Russia) use inputted information for further model training. This means:
Your test results, symptoms, diagnoses, and even your full name can end up in training datasets.• Over time, this information may “surface” in responses to other users — anonymized, but still potentially recognizable.• In the event of a data leak or system breach (which has already happened in the past), your personal medical details may become publicly accessible.Gorelkin noted: “Any data you send to an AI system can potentially be used to improve its algorithms. If the information should never become public, it is better not to upload it at all.”
What to Do Instead of Using AI
Use only official, licensed medical services with encryption to interpret test results (Invitro, Gemotest, SberHealth with government-authorized login). Ask AI only anonymized questions: for example, “What can elevated ALT indicate?” instead of “Interpret my test results from November 20.” Remember: no AI in 2025 has medical training or legal responsibility for incorrect advice.In Short
Over six months, the number of Russians uploading their medical tests to AI and asking “what to do” has increased eightfold. Expert Alexey Gorelkin warns that such data is almost certainly used for model training and may eventually become public. A neural network is not a doctor, and your medical information is confidential — it should not be entrusted to AI. Anonymized questions are acceptable; personal test results are absolutely not.
Follow NEWS.am Tech on Facebook and Twitter