CISA leads tabletop exercise focusing on hacked AI

The Cybersecurity and Infrastructure Security Agency said that last week’s inaugural tabletop with the private sector, coordinated by the Joint Cyber Defense Collaborative at the Microsoft facility in Reston, Virginia, supports the development of a cross-sector AI Security Incident Collaboration Playbook to be published by end of the year.

WHY IT MATTERS

A dedicated planning effort within JCDC, CISA’s public-private partnership model driving collaboration on preparedness among AI providers, security vendors and critical infrastructure owners and operators, addresses risks, threats, vulnerabilities and mitigations concerning AI-enabled systems in national critical infrastructure, the agency said in a statement June 14.

More than 50 organizations participated in the recent four-hour preparedness exercises sharing their strategies for adopting AI securely to protect critical infrastructure from emerging threats and practicing collaborative response.

“Simulating adversarial threats against AI systems in a controlled setting is an invaluable training ground to equip security teams with an understanding of the vulnerabilities and threats that exist today,” said Chris Sestito, chief executive officer and cofounder of HiddenLayer.

Other technology firms at the table were Amazon Web Services, Cisco, IBM, Microsoft, NVIDIA, OpenAI, Palantir, Palo Alto Networks, Protect AI and more leading vendors. They were joined by the Federal Bureau of Investigation, National Security Agency, Office of the Director for National Intelligence, Department of Defense and Department of Justice. 

“This exercise marks another step in our collective commitment to reducing the risks posed by AI,” Easterly said in the statement.

“As the adoption of AI has expanded, we’ve seen a similar growth in complexity in the cyber threat environment,” added Sandy Reback, vice president of public policy and government affairs, at Palo Alto Networks. 

“Public-private collaborations on critical exercises like this will better protect our digital way of life.”

For Bryan Vorndran, FBI’s Cyber Division assistant director, the exercise demonstrated the agency’s commitment to partnerships, he said.

According to CISA, the need for secure-by-design approaches in developing AI products was also an important theme beyond incident-response collaboration and practice at the tabletop exercise.

“These collaborations benefit our efforts of safely developing and deploying AI technology,” Matt Knight, head of security at OpenAI, acknowledged in the statement.

JCDC is planning a second exercise in 2024 that will incorporate vulnerabilities related to system integrators in U.S. critical infrastructure, which enable interoperability when implementing AI technologies into existing systems. The AI integrators help organizations adopt AI and create larger AI systems.

“With critical infrastructure facing increasingly severe attacks and the rise of AI threats, early preparedness and routine testing is more important than ever to reduce any collateral damage,” Troy Bettencourt, global partner and head of IBM X-Force, noted in the agency announcement.

The AI Security Incident Collaboration Playbook, to come out of CISA’s tabletop exercises with the private sector at the end of the year, is intended to facilitate AI security-incident-response coordination efforts between government, industry and global partners, the agency said.

Omar Santos, who leads security and trust at Cisco, called the playbook “a well-needed initiative” that will “serve as a great resource for coordinating AI security incidents among industry peers and global partners.”

THE LARGER TREND

As part of its mission launched two years ago, the JCDC said it is working to decrease the likelihood and impact of AI-related threats and vulnerabilities to critical infrastructure providers on its website.

Emerging technologies always present a ripe opportunity for tabletop exercises, and experts agree that when protecting critical infrastructure, the government is a key partner. 

“There are laws that define this relationship, specifically the National Defense Authorization Act. This codifies the critical infrastructure relationship between the Federal government, through a Sector Risk Management Agency and the [critical infrastructure],” Erik Decker, Intermountain Health CISO and co-chair of the HHS 405(d) Task Group, noted when sharing tips on provider-organization cybersecurity tabletop exercises with Healthcare IT News.

ON THE RECORD

“At OpenAI, we firmly believe that security is a team sport. It thrives on collaboration and benefits immensely from transparency,” Knight said in a statement. “This initiative not only strengthens our defenses but also fosters a community dedicated to collective security advancements, which includes realizing the benefits of using AI tools for cyber defense.”

“As we enter a new AI Landscape, security is critical, and collaboration with industry and government partners is crucial to developing an effective and coordinated response to security incidents,” added Bret Arsenault, Microsoft’s corporate vice president and chief cybersecurity advisor.

“Practicing response scenarios and simulations like today’s AI-focused tabletop exercise drive learning and sharing that will help strengthen cyber resilience across the board.”

Andrea Fox is senior editor of Healthcare IT News.
Email: afox@himss.org
Healthcare IT News is a HIMSS Media publication.

The HIMSS AI in Healthcare Forum is scheduled to take place September 5-6 in Boston. Learn more and register.