How AI in Medical Imaging Could Expose Your Private Health Data – and What to Do

Artificial intelligence is transforming medical imaging. Algorithms now help radiologists spot tumors, measure organ changes, and prioritize urgent cases. But this progress comes with a less visible cost: the privacy of your medical data. A recent report from the Radiological Society of North America (RSNA) warns that the same AI tools that improve diagnosis can accidentally – or deliberately – expose sensitive health information. For patients, understanding these risks is the first step toward protecting their own records.

What Happened

The RSNA report, published in May 2026, examines how AI systems used in radiology create new avenues for privacy breaches. It highlights three specific threats: the re-identification of supposedly anonymous images, the unauthorized secondary use of patient scans, and the lack of clear consent when medical images are used to train AI models.

Medical imaging files – X-rays, CT scans, MRIs – contain far more than a picture of an organ. They often include metadata such as patient names, dates, and sometimes embedded genetic or anatomical markers that can be linked back to an individual. Even after metadata is stripped, researchers have shown that facial recognition techniques or unique bone structures can re-identify patients.

The RSNA report also notes that many patients are unaware that their de-identified scans are being used to train commercial AI systems. Consent forms rarely mention this secondary use, and patients may assume their images are only seen by their own care team.

Why It Matters

The shift to AI-assisted radiology is accelerating. Major hospitals and imaging centers now license AI tools from third-party vendors, which means patient data often leaves the healthcare network. The privacy safeguards patients expect – such as HIPAA protection in the United States – may not extend to how the vendor handles the data afterward.

Re-identification is not theoretical. In 2019, researchers demonstrated they could re-identify participants from a public medical imaging dataset using face-matching software. Similar techniques are improving. If an insurer, employer, or data broker obtains a re-identified image, it could be used to deny coverage, make hiring decisions, or target advertising based on a person’s health status.

Moreover, AI models trained on biased or incomplete datasets can produce less accurate results for certain populations. But the privacy risk is immediate: once your imaging data is shared for AI development, you lose control over who sees it and how it is used.

The RSNA report argues that the current system lacks transparency. Patients are rarely asked whether their images may be used for AI training, and even when they are asked, the language is often vague. This gap between expectation and reality undermines trust.

What Readers Can Do

While patients cannot single-handedly change hospital policies, there are practical steps you can take to protect your medical imaging privacy.

Ask questions before your scan. When scheduling or arriving for an imaging exam, ask the facility how your images will be stored, who has access, and whether they are used for AI development. Request a copy of the consent form and read it carefully. If the language is broad – for example, “your data may be used for quality improvement” – ask for clarification.

Opt for de-identified data sharing if you have a choice. Some institutions allow patients to choose whether their anonymized images can be used for research or commercial AI training. You have the right to say no. If the facility does not offer this option, consider requesting a written assurance that your data will not be shared outside your care team.

Check your provider’s privacy policy. Larger healthcare systems often publish privacy policies online. Look for sections on “data sharing,” “third-party vendors,” or “research use.” If the policy is unclear, call the privacy office.

Use patient portals cautiously. Many portals now include an option to consent to research. Before clicking yes, understand what data is included – imaging, lab results, clinical notes – and whether the data is truly anonymized.

Stay informed. The RSNA report is a good starting point, but privacy risks evolve. Follow credible sources such as the Electronic Frontier Foundation or the American Medical Informatics Association for updates on health data privacy.

Sources

  • Radiological Society of North America, “Medical Imaging AI Opens a Pandora’s Box of Privacy-Related Risks,” May 2026.
  • Schwartz et al., “Re-identification of Medical Images Using Facial Recognition,” Nature Communications, 2019.
  • HIPAA Privacy Rule, U.S. Department of Health and Human Services.