Feds Prep for AI Cyberattacks, Ban DeepSeek

Cyberattacks using artificial intelligence jumped 89 percent in 2025 compared with 2024, and federal security officers are preparing for more. Mike Duffy, the acting federal CISO, said the Federal Chief Information Security Officer's Council will bring agencies together in early November for a governmentwide tabletop exercise on AI-enabled attacks. It’s a practice run for the real thing.

Who Is Exposed

Duffy said the exercise will help agencies assess the capabilities, processes and personnel they need to respond to cyber events. The council’s second all-government CISO-level tabletop follows its first, held the previous year and focused on broader readiness and resilience. This time, the focus is narrower and more urgent. At the ACT-IAC Cybersecurity Summit on Oct. 14, Duffy said the question has shifted from what one agency should do to what the threat means for all of them at once.

The threat picture is getting worse for people who depend on these systems. CrowdStrike’s 2026 report found that hackers using AI-accelerated phishing and automated reconnaissance were shortening the time between initial access and impact. The report said AI attacks had elevated less sophisticated threat actors and amplified the most advanced ones. The barrier to entry is falling. Ordinary citizens whose tax records, health data and benefits sit in federal databases absorb the fallout when a breach succeeds.

The Policy Gap

Rubrik’s public sector chief technology officer, Travis Rosiek, argued in a Federal News Network commentary that agencies can’t count on keeping intruders out. He wrote that disaster recovery plans built for floods and power outages don’t adequately address cyberattacks, and pointed to the self-replicating Shai-Hulud worm, which hackers deployed into developer pipelines in 2025. The worm bypassed perimeter defenses, harvested cloud credentials and moved into both private and public networks. Rosiek urged agencies to design systems that can isolate breached segments while keeping core functions running. He also called for federal, state and defense networks to meet the standards already applied to critical infrastructure operators.

Rosiek also cited the Office of the Director of National Intelligence’s 2026 Annual Threat Assessment. It said adversaries were exploiting supply chains to compromise software source code and pre-position themselves inside federal networks. The assessment said their objective was persistent access to shut down critical services, disable power grids and disrupt military readiness. When the stakes include the lights going out for families, resilience stops being a bureaucratic exercise.

Duffy said the CISO Council had revived its AI working group, focused on security governance, and started a post-quantum cryptography working group the previous month. He described the post-quantum effort as a multiyear project requiring expertise and workforce migration. The National Institute of Standards and Technology and the Cybersecurity and Infrastructure Security Agency are among the organizations working on it.

Official Response

Duffy said the Trump administration released a national cyber strategy in March. The Office of Management and Budget has focused on policies, procedures and capabilities to defend against attacks, including changes to cyber logging architecture policy. He said the Cybersecurity and Infrastructure Security Agency had changed how it manages the known exploited vulnerabilities catalog, prioritizing fixes by likely exploitation and potential mission impact. Agencies are applying those changes while they work under a binding operational directive to secure and modernize endpoint devices.

A sprint inside the CISO Council is assessing how much progress agencies can make with existing resources before the fiscal year ends, including removing outdated edge devices and infrastructure. Duffy said the long-term goal is sustained operational resilience: repeated assessments of which systems need attention. That’s sensible, though it only works if agencies have the staff and money to act on what they find.

The Department of Homeland Security took a separate step. It ordered federal civilian agencies to remove AI developed by China’s DeepSeek from their systems and barred future purchases of its products and services, including third-party services built on the company’s technology. The DHS notice said the company’s models “present multiple threat vectors” and that use of its products “may result in the transfer of sensitive data to the [People’s Republic of China] if used by US government employees or vendors.” DHS acted under the Federal Acquisition Supply Chain Security Act of 2018, after reviewing a recommendation from the Federal Acquisition Security Council. The notice didn’t detail the security concerns or set a deadline for agencies to comply.

That silence on details matters. DeepSeek, founded by Liang Wenfeng, offers cheaper alternatives to products from Anthropic and OpenAI. Its performance had lagged leading U.S. models, though the gap narrowed over the previous year by some measures. The Commerce Department banned it from government devices last year, the U.S. Navy told its members not to use it, and New York state barred it from government networks, citing “serious concerns” about data privacy and censorship.

Data Governance Debate

Taka Ariga, vice president for data policy and impact at the Data Foundation, used a separate Federal News Network commentary to call for a national data strategy. He wrote that the United Kingdom published a National Data Strategy in 2020, while the United States has no equivalent. Data.gov launched in 2009 and provides public access to many federal, state and local open datasets. The Foundations for Evidence-Based Policymaking Act, signed by President Donald Trump in 2019, requires a standard application process for researchers seeking restricted federal data. The Federal Data Strategy, released that same year, drew on input from industry and researchers.

Ariga argued that generative AI has made public data both more valuable and more sensitive. States and localities are building their own data capacity, while private and philanthropic money is flowing into projects that combine public and private information. Expectations about privacy are shifting as data is combined across sectors. The Data Foundation has launched a community-led effort with a living document meant to gather feedback and critique. Ariga said it isn’t meant to replace the existing federal framework but to extend it across society. Drafts will be posted at NationalDataStrategy.us.

Why This Matters:

The people most affected by a successful attack on federal systems aren’t the officials running tabletop exercises. They’re the veterans waiting on benefits, taxpayers whose returns sit in government databases, and families who depend on the power grid and public health services adversaries have reportedly targeted. Rosiek’s warning that disaster recovery plans were built for floods, not hackers, points to a gap that falls hardest on those with the fewest options when systems go dark. Preparation is a public responsibility, and it deserves public money, public standards and public accountability. Whether the November exercise produces funded, binding changes or simply another report will show whether the government treats AI-driven threats as the structural risk they are. The documents here show a strategy in progress, not a finished defense.