Navigating the shift from volume to value in healthcare
“What we are seeing is a real challenge,” says Brian Anderson, M.D., CHAI’s president and CEO.
The Coalition for Health AI (CHAI) recently launched the Health AI Cybersecurity Work Group to help healthcare organizations address cybersecurity risks posed by advanced AI models. The group plans to develop practical, peer-led guidance to help health systems, payers, and health technology companies prepare for potential threats and use frontier AI models to strengthen cyber defenses.
Speaking at a Healthcare Summit hosted by cybersecurity company Rubrik, Brian Anderson, M.D., CHAI’s president and CEO, discussed in more detail some of the issues the work group needs to address. Anderson warned that advanced AI could accelerate cyberattacks and data theft, potentially beyond the speed of human response.
Healthcare Innovation spoke with Brian Anderson in more detail.
Anderson said the initiative responds to growing concerns about advanced AI models’ ability to conduct sophisticated cyberattacks at unprecedented speed. Healthcare organizations face particular risks because cybercriminals frequently target the sector.
“What we are seeing is a real challenge,” Anderson said, pointing to the growing capabilities of frontier AI models and the need for healthcare-specific cybersecurity strategies.
One major concern is what Anderson described as “time compression.” Tasks that once took human attackers days or weeks, including penetrating systems, maintaining access, and extracting sensitive information, may increasingly be completed by AI in milliseconds.
“We are going to need to have AI operate in a defensive manner in what is likely a semi-autonomous way,” Anderson said, arguing that AI will need to help protect health systems and health plans at a speed human teams cannot match.
The initiative will also address security risks associated with agentic AI, which can perform tasks and coordinate complex workflows with limited human intervention. Anderson emphasized the importance of controlling these systems’ identities and permissions, including which data and systems they can access and which other agents they can communicate with.
The working group will develop its guidance with input from healthcare organizations, technology companies and frontier AI developers, including Anthropic and Rubrik. Anderson said bringing these groups together and testing emerging models against defensive strategies will help organizations adapt as threats evolve.
CHAI, founded in 2021 to advance responsible AI development and use in healthcare, aims to help organizations benefit from AI while protecting sensitive patient information and critical infrastructure.
Ultimately, Anderson said, the goal is to create “a safer, more trusted healthcare ecosystem” in which AI supports healthier lives while helping keep healthcare providers and payers secure. He added that the initiative is open to additional participants and that its resources are intended to be available for operational use as quickly as possible.
Pietje Kobus-McAllister has an international background and experience in content management and editing. She studied journalism in the Netherlands and Communications and Creative Nonfiction in the U.S. Pietje joined Healthcare Innovation in January 2024.
About the Author
