Forging the Guardrails: How Datavant is De-Risking AI for Healthcare

NEW YORK, NY – June 03, 2026 – In the rapidly advancing world of artificial intelligence, the line between powerful tool and autonomous partner is blurring. For industries like finance and tech, this evolution promises unprecedented efficiency. But in healthcare, where decisions carry the weight of human life, it introduces a new frontier of risk. The central challenge is no longer just about building smarter AI, but about building trustworthy AI. Addressing this head-on, Datavant, a cornerstone of healthcare data collaboration, has announced it is joining the AIUC-1 Consortium, a cross-industry group tasked with writing the safety rulebook for this next wave of autonomous, or 'agentic,' AI.

This move signals a critical shift from theoretical ethics discussions to the creation of tangible, testable standards. By lending its deep expertise in the complex world of health data, Datavant is helping to ensure that as AI agents begin to operate in hospitals and clinics, they do so on a foundation of absolute trust, security, and reliability.

From Data Plumbing to AI Architecture

For years, Datavant has operated as the trusted, neutral infrastructure of the healthcare data ecosystem. The company doesn't own or sell data; instead, it provides the sophisticated 'plumbing' that allows payers, providers, and life sciences companies to connect disparate datasets while rigorously protecting patient privacy. Through advanced, HIPAA-compliant tokenization and de-identification technologies, it enables the secure exchange of over 60 million health records, underpinning critical research and operational insights. With an annual investment of over $40 million in its security and compliance infrastructure, and certifications like SOC 2, HITRUST, and FedRAMP, the company has built its reputation on being an unimpeachable steward of sensitive information.

Joining the AIUC-1 Consortium is a natural, if ambitious, progression of this role. Having mastered the art of building trust in data exchange, the company is now extending that expertise to the AI systems that will increasingly act upon that data. It’s a strategic move from ensuring the integrity of data at rest and in transit to guaranteeing the integrity of AI-driven actions.

“By collaborating with the Artificial Intelligence Underwriting Company and leaders across industry, we are ensuring that as the speed of AI expansion increases, healthcare-specific risk controls scale at the same rate,” said Dan Walsh, Datavant's Chief Information Security Officer. This statement encapsulates the company’s proactive stance: instead of waiting for AI-related incidents to force a reaction, Datavant is helping to build the necessary guardrails in advance, architecting a framework for safety before these powerful new tools are widely deployed in high-stakes clinical environments.

Defining the 'Agent': The New Frontier of AI Governance

To grasp the significance of this initiative, it's crucial to understand the technology at its core. 'Agentic AI' represents a leap beyond the generative AI many have become familiar with. While a large language model can answer a question or draft a document, an AI agent can take that output and act on it to achieve a multi-step goal with limited human oversight. It can interact with other software, access databases, and make decisions to accomplish a task, whether it's optimizing a hospital's supply chain, scheduling a complex series of patient appointments, or even analyzing clinical trial data to identify candidates.

This autonomy is precisely what makes agentic AI so powerful, but also what makes robust governance so non-negotiable. Recognizing this, the AIUC-1 Consortium has developed what it calls “the world’s first AI agent standard.” It’s not another abstract whitepaper on ethics, but a practical, testable certification framework built on six core pillars: Data & Privacy, Security, Safety, Reliability, Accountability, and Society. The standard is designed to be a clear signal of trust, enabling organizations to confidently adopt AI technology from certified vendors.

The consortium’s roster lends significant weight to the effort. With technical contributions from institutions like MITRE, which integrates its adversarial threat knowledge base, and Stanford's Trustworthy AI Research Lab, the AIUC-1 standard is grounded in deep, multi-disciplinary expertise. It aims to operationalize high-level principles from frameworks like the NIST AI Risk Management Framework and the EU AI Act, focusing specifically on the observable behavior of AI agents in real-world production environments.

Tailoring the Blueprint for Medicine's Unique Risks

While a general standard for AI safety is a major step forward, healthcare operates under a unique set of pressures that demand a more specialized approach. The potential for harm is not merely financial or reputational; it is physical and deeply personal. A generic framework might not adequately account for the stringent requirements of HIPAA, the nuances of protecting Protected Health Information (PHI), or the life-or-death reliability needed for clinical decision support.

This is where Datavant’s contribution becomes invaluable. The company’s role within the consortium is to embed this deep domain knowledge into the DNA of the AIUC-1 standard, adapting its controls for the specific challenges of healthcare. For example, under the 'Reliability' pillar, this could mean defining specific thresholds to prevent an AI from 'hallucinating' or fabricating information when analyzing a patient's medical history. Under the 'Security' pillar, it could involve creating controls to ensure an AI agent used for administrative billing cannot access sensitive clinical notes.

“AIUC is building the trust infrastructure necessary for AI agents to be deployed safely across the enterprise, and there is no sector where trust is more paramount than healthcare,” noted Rajiv Dattani, Co-founder of AIUC. “Datavant’s deep experience managing data in the field is invaluable.” By helping to ensure the standard is as effective in a hospital as it is in a financial institution, Datavant is directly tackling one of the biggest barriers to AI adoption in medicine: the trust deficit.

Building a Coalition for Trustworthy Innovation

Ultimately, the story of Datavant and the AIUC-1 Consortium is a story about the power of collaboration in the face of transformative, and potentially disruptive, technological change. No single company, no matter how influential, can solve the challenge of AI safety alone. The consortium model, which unites over 200 leaders from security, technology, healthcare, and finance, acknowledges that building a universal 'trust infrastructure' requires a diverse coalition of experts.

This collective effort is creating a practical pathway for organizations to move from cautious curiosity about agentic AI to confident adoption. For a healthcare system grappling with burnout, rising costs, and operational inefficiencies, the promise of autonomous AI is immense. It could free up clinicians from administrative burdens, accelerate life-saving research, and personalize patient care at an unprecedented scale.

However, realizing that promise depends entirely on the unglamorous, foundational work of building robust guardrails. By contributing its expertise to a concrete and technically grounded standard, Datavant is not just solidifying its position as a market leader; it is helping to lay the very groundwork upon which a safer, more intelligent, and more resilient future for healthcare can be built.