Army Expands AI Power as GSA Sets New Rules

The Army is developing AI agents that can detect and respond autonomously to cyber threats. Meanwhile, new General Services Administration rules let federal contracting officers add AI safeguards to government contracts. Together, the efforts give federal agencies a larger role in setting terms for AI use and expanding its place in government systems and cyber operations.

Federal Rules for AI Purchases

The GSA class deviation, GSAR 552.239-7001, took effect Monday, October 19. Its title is “basic safeguarding of data within large language model artificial intelligence systems.” Contracting officers can use the clause now when a large language model is a material feature of a purchased product or service.

The clause will appear in the next multiple-award schedule refresh, refresh 34. Contractors will have 30 days after that refresh, and implementation is expected to begin three months later. The clause will also go through a rulemaking process, giving interested parties another chance to comment.

Emily Murphy, Senior Fellow at the George Mason University Baroni Center for Government Contracting and former GSA administrator, said the first version of the clause came out in March. It gave the government extensive rights, set requirements on where AI was developed and used, assigned the government ownership of processes involving AI, and included neutrality requirements. GSA withdrew it after receiving comments, issued a second version, then gathered further comments and held an industry listening session before producing the third version.

“They've got a tough challenge because they've got an administration that has overall said, we don't want to regulate superintelligence,” Murphy said. She said the government must protect taxpayer interests as an AI user, ensure data gets used appropriately and avoid creating vulnerabilities.

Murphy described the latest clause as more closely tied to NIST standards, with a focus on objective measures and benchmarking. It includes a 25% cap on a contractor's extraction liability if the contractor must remove AI from a system following termination for default. The clause also addresses bringing AI into government systems and removing it, neutrality, and differences between open-source and open-weight models.

Contractors Face Unsettled Boundaries

The clause applies when an LLM is a material feature, but Murphy said the boundary remains ambiguous. Earlier versions exempted productivity tools; the latest clause could cover spreadsheets with built-in AI or Zoom's AI noise suppression if those uses count as material. Back-office administrative functions don't count. Murphy said contractors should document how they use LLMs.

The clause gives contractors stronger intellectual-property protections, streamlines reporting and requires reports of an LLM issue within 72 hours. Existing reporting under FedRAMP or CISA counts. Murphy raised concerns about the clause overriding end-user licensing agreements for commercial software, and about contractors passing requirements to suppliers when they may lack leverage to negotiate.

Murphy said decisions about material use could depend on contracting-officer discretion or negotiations between agencies and contractors, so uniform implementation is unlikely at the outset. Implementation will take time and continue to be refined. GSA held two comment periods and listening sessions during the preceding six months, and Murphy said it would keep collecting feedback and lessons before finalizing its approach.

Army Builds an Agentic Cyber Team

Army Cyber Command launched Task Force Lexington earlier in 2026 to develop AI agents for cyber functions. The task force first created four AI-agent work roles and has since expanded them. Lt. Gen. Christopher Eubank, who leads the command, said it had built an entire agentic cyber protection team.

“Since August, the biggest thing we have done is complete the cyber protection team. We now have all the work roles inside of a cyber protection team mapped out at this point, and learning their trade,” Eubank told reporters at the Association of the U.S. Army's annual conference in Washington. The roles include analytic support officers, data engineers, network analysts and host analysts.

The task force trained the agents, ran them in a sandbox, retrained them and then let them work on the Army's portion of the Defense Information Network inside its big data platform. Eubank said the agents were learning to identify compromises and had been “passed on to the joint force to do defensive work on behalf of the [Defense Department].” He added, “What we've built, we have given to Lt. Gen. Stanton at the Department of Defense Cyber Defense Command.”

Eubank said strict guardrails kept the Army's agents from straying. He described them as being treated “just like a human operator,” with limited permissions and humans checking their work every day. The Army is also set to award contracts to two companies for Project Griffin, a pilot program to develop agents that can autonomously respond to cyberattacks. Brandon Pugh, the Army's principal cyber advisor, said the effort would pursue autonomous action within parameters set by a human operator.