Garbarino report warns of potential AI-driven attacks

WASHINGTON — A new report released by Long Island Rep. Andrew Garbarino sounds alarms about the rise of artificial intelligence and its potential for malicious use in attacks against the United States.

Overall, the 103-page report by the Committee on Homeland Security, chaired by Garbarino (R-Bayport), outlines an ever-evolving threat landscape that is "both familiar and unknown." As "adversaries have changed" over the 25 years since the Sept. 11, 2001, attacks, the report states, so have advances in technologies that continually complicate U.S. security practice.

"This no-fail mission requires persistent vigilance and adaptation to keep pace with the contemporary threat environment," Garbarino writes in the report’s preface.

While the report also focuses on myriad security improvements over 25 years, among its most sobering discussions is the rise of artificial intelligence, and the prospect of its uses by adversaries domestically and abroad to harm Americans.

In doing so, the report’s discussion goes beyond earlier reports recounting headline-grabbing cyber-related campaigns carried out in recent years by nation states or others, some "with the resources, patience, and capabilities of intelligence services and military organizations."

Targets can and have included communications networks, transportation systems, financial services, energy infrastructure, water systems, healthcare facilities and government services. Cyber enemies, the report states, "steal sensitive information, disrupt operations, compromise trusted vendors and service providers, abuse legitimate administrative tools and remain inside networks for extended periods, often without being detected."

The report underscores that AI also is part of an arsenal to boost or enhance U.S. security because it can help defenders analyze large amounts of data, identify suspicious activity, prioritize vulnerabilities, support compliance and automate repetitive tasks.

"But the same capabilities can also assist malicious actors," the committee warns. "AI can improve reconnaissance, facilitate social engineering, accelerate vulnerability research, generate or modify malicious code and allow threat actors to conduct operations with greater speed and scale."

The report notes that "lone-wolf" or lone-actor plots by individuals not inspired by others or organized groups have emerged as increasingly sharper concerns, requiring some shifting in recent years from organization-focused counterterrorism, to "behavior-centric prevention." 

"Developments in artificial intelligence have made it easier for unsophisticated threat actors to conduct complex attacks against the United States that would have previously required far greater resources and planning," the report warns.

As advanced AI models become more capable of performing multistep tasks with limited human direction, "the distinction between a tool that assists an operator and a system capable of executing substantial portions of an operation autonomously will continue to narrow," the report states.

The committee’s top Democrat, Rep. Bennie Thompson of Mississippi, noted in a statement Friday that the Trump Administration has been carrying out personnel and program cuts at the Cybersecurity and Infrastructure Security Agency (CISA). And the Trump administration’s fiscal 2027 budget plan proposes removing another 900 positions from the agency, Thompson said.

"Cutting national security staff, undermining cybersecurity ... and withholding counterterrorism funding create dangerous security gaps," Thompson said. "After 9/11, we worked across party lines to strengthen our homeland security. Today, those protections are being weakened as threats continue to evolve."