Public Health

Viz.ai vs. HIPAA: Breach Data Exposes AI Security Gaps

Listen to this article · 9 min listen

Everyone talks about healthcare AI with grand promises of innovation, but the reality on the ground is that our digital health infrastructure is a sitting duck for cyberattacks. Compliance officers and security leaders, the ones actually responsible for safeguarding patient data, know you have to weigh the promise of AI efficiency against the hard evidence of ongoing security vulnerabilities. You can see this conflict play out when you examine long-term HIPAA enforcement trends against the voluntary security frameworks that vendors claim to adopt.

The HHS breach portal gives us over fifteen years of public HIPAA enforcement data, one of the best records we have for looking at vendors. After digging through this extensive record which covers 2009 through 2026, firms like Medcrypt have concluded that the data breach trend is flat-out not improving. This is a big problem for any talk about AI regulatory compliance, because all the optimism about securely plugging in AI tools rests on the quiet assumption that our underlying data is getting safer. But when you put that stubborn breach record up against the documented NIST Cybersecurity Framework posture of a major imaging AI company like Viz.ai, a huge gap opens up between having a framework and seeing fewer incidents. So for compliance and audit teams, the real question is: what does that long breach record actually tell you, and where should you focus first?

The Unflattering Trend from the HHS Breach Portal

Looking at the full HHS breach portal data from 2009 to 2023, Medcrypt’s analysis makes one thing very clear: things aren’t getting better. That conclusion, based on more than a decade of public reports, blows a hole in the assumption baked into a lot of healthcare AI compliance pitches, that the data environment is somehow improving on its own. The constant stream of reported incidents shows that for all the talk about awareness and new security tools, the healthcare sector is still failing at basic data protection. Anyone tracking HIPAA enforcement data will see this analysis and recognize the pattern. The Medcrypt analysis of HHS breach portal data is a cold reminder that digital health is always under attack, and any AI you deploy has to live in that dangerous world.

Healthcare AI faces some big problems because of this. When AI models get plugged into clinical workflows and start chewing on massive amounts of protected health information (PHI), data integrity and security become everything. We talk a lot about algorithmic drift in terms of clinical accuracy, but it has a security side too: if the data feeding the AI gets compromised, the model’s outputs become untrustworthy and the whole system is undermined. Medcrypt’s findings force us to go back and question the basic security assumptions for new AI deployments. You can’t just drop a new AI tool into the hospital network. The environment has to be provably secure, and the current breach data shows that’s a huge uphill battle.

Viz.ai’s NIST Cybersecurity Framework Posture Against the Backdrop of Breaches

On the other side of this grim breach data, you have healthcare AI companies like the imaging AI firm Viz.ai, who point to their adoption of strong security frameworks. Viz.ai, for example, includes the NIST Cybersecurity Framework in its audit standards. The NIST Cybersecurity Framework, detailed in documents like NIST SP 800-66r2, is a voluntary but thorough set of guidelines for managing cybersecurity risk, as you can see in the NIST Cybersecurity Framework official documentation. It gives companies a structured way to handle threats, identify, protect, detect, respond, and recover, and adopting it is generally seen as a sign of a mature security program. For a compliance officer, aligning with a framework like this is just table stakes for demonstrating you’re doing your due diligence on data security.

But when you put the breach trends from Medcrypt next to Viz.ai’s documented NIST posture, you see a big disconnect. The real question is whether Viz.ai has actually bent the curve on incident counts for companies that use it. All the public material from Viz.ai shows they stick to the NIST Cybersecurity Framework, but there’s no hard evidence that having this framework has made any difference to breach outcomes for companies like them. Frameworks are obviously valuable for setting a baseline of security practices. But the HHS portal data shows that simply adopting a framework, without constantly and rigorously validating it against real-world threats, isn’t enough to stop the breaches. That gap, between the voluntary framework and the actual incident counts, is what HIPAA enforcers and AI compliance teams need to scrutinize.

The Gap Between Frameworks and Real-World Incidents

The whole story here is the gap between the strong, voluntary security frameworks that leading AI healthcare companies adopt and the unending trend of data breaches recorded by the HHS. Any compliance audit looking at this situation would have to look past the simple claim of “framework adoption.” The point is straightforward: Medcrypt’s read of HHS breach portal data shows incident counts that haven’t been bent by a recorded NIST Cybersecurity Framework posture like the one Viz.ai has. This is an important point for compliance officers. Having a framework in your audit standards helps set a baseline, but it doesn’t automatically mean you’re stopping more breaches.

Think about this in the context of all the AI healthcare regulation updates happening around 2026. As regulators like the FDA and professional groups like the AMA continue to tighten oversight for AI, with new things like the ECRI AI healthcare hazard 2026 rankings and AMA AI healthcare oversight 2026 guidelines, the real-world effectiveness of today’s security measures is going to be a huge point of debate. If existing frameworks, even good ones like NIST, aren’t visibly cutting down the number of breaches, then just relying on them to secure AI systems isn’t going to be good enough. The industry has to stop just checking the box on frameworks and start showing that they lead to measurable improvements in security. A smart observer should watch to see if these framework adoption claims ever get tied back to the breach portal counts, and if any company can show improvement the data actually backs up.

Where a Compliance Audit Would Look First

With the persistent breach trends and the clear gap between adopting a framework and reducing incidents, any good compliance audit will prioritize verification over declaration. For compliance officers, relying on a vendor’s claim of NIST Cybersecurity Framework adoption without independently verifying it against real-world performance metrics would be a major oversight. The audit has to start by looking at direct evidence of security effectiveness. It’s about getting past paper compliance to see demonstrable, operational security.

  • Incident Response and Remediation Records: An auditor is going to demand the full story on any past incidents, asking for detailed records of how they were handled, the timelines, root cause analyses, and proof that your corrective actions actually worked. The fact that you can’t find audited figures on incident reduction or remediation costs for any framework holder is a huge red flag that an audit would jump on.
  • Continuous Monitoring and Testing: They’ll want proof of ongoing vulnerability assessments, penetration testing, and real-time security monitoring. A framework posture that just sits on a shelf is worthless without dynamic validation against constantly evolving threats.
  • Data Flow Mapping and Access Controls: For AI systems, understanding the whole data lifecycle is non-negotiable, so auditors will pore over data flow maps, encryption protocols, and granular access controls, especially for PHI. This is straight out of HIPAA’s playbook for protecting electronic protected health information (ePHI) HHS HIPAA guidance.
  • Evidence of Improvement Tied to Frameworks: An audit will look for a causal link: how did implementing the NIST Cybersecurity Framework specifically reduce vulnerabilities or improve your threat response, and can you show it with a lower incident count or faster remediation times? Without that verifiable tie, framework adoption is just an aspiration, not a proven defense.

The message for compliance officers is clear: treat framework adoption claims as unproven until they are tied directly to incident counts and show improvement that the data verifies. You have two checkable anchors, the breach portal counts and the documented framework posture, but the real story lies in the distance between them.

Frequently Asked Questions

What does the HHS breach portal data, analyzed by Medcrypt, reveal about the trend of data breaches in healthcare?

Medcrypt’s analysis of HHS breach portal data from 2009 to 2023 indicates that the trend of data breaches in healthcare is not improving. This suggests that despite growing awareness and evolving security measures, the healthcare sector continues to struggle with fundamental data protection, challenging assumptions of an increasingly robust and safer data layer for AI integration.

How does the observed breach trend compare with the adoption of security frameworks like the NIST Cybersecurity Framework by companies such as Viz.ai?

There is a critical gap between the observed, persistent breach counts and the voluntary adoption of robust security frameworks like the NIST Cybersecurity Framework. While companies like Viz.ai adhere to such frameworks, the long-term breach record does not show that framework adoption alone has demonstrably reduced incident counts for holders.

What is the implication of this gap for compliance and audit teams evaluating AI healthcare regulatory compliance?

The gap between voluntary security frameworks and observed incident counts suggests that framework adoption alone may not be sufficient to stem the tide of breaches without rigorous, continuous validation against real-world threats. Compliance and audit teams should focus their scrutiny on this disparity, questioning what a long breach record truly reveals when set against a documented framework posture.

Does the article suggest that adopting security frameworks like the NIST Cybersecurity Framework is ineffective?

No, the article does not diminish the value of adopting security frameworks; it states they are essential for establishing a baseline of security practices. However, it highlights that framework adoption alone, without continuous validation against real-world threats, may not be sufficient to significantly reduce the number of breaches.

Share
Was this article helpful?

Editorial Team

The Editorial Team at Healthcare AI Compliance Watch is a group of writers and editors who cover health, workplace wellness, and everyday care. They research each story and publish under a shared desk byline — experts in their fields, working as one team.