Army Advances Project Griffin AI Cyber Response Awards

Army Cyber Command (ARCYBER) is advancing Project Griffin, a pilot focused on AI agents that can detect potential cyber threats and take action within parameters set by human operators. The initiative sits within a broader Army effort to build an agent workforce alongside personnel for cyber operations, network security, and defense. Army officials have described a deliberate approach: agents are trained for defined work roles, monitored by people, and evaluated before being given greater autonomy. The supplied reporting says Project Griffin’s agents will undergo evaluation and sandbox testing with human oversight before any use with Army network data; it does not establish that they will immediately operate on government networks.

The Army selected two companies from a field narrowed to 10, following more than 100 industry submissions. That level of participation signals strong vendor interest in military applications for agentic AI, but the public information provided does not identify the selected firms or disclose award values. A planned announcement date of October 16, 2026, has passed, yet the available signals do not confirm whether awards were announced. For contractors, this means Project Griffin is a meaningful market indicator, but not a sufficiently documented award notice for attribution, pricing analysis, or subcontracting outreach tied to named recipients.

The project’s stated focus is autonomous response under human-defined constraints—not unrestricted machine action. Army Principal Cyber Advisor Brandon Pugh described the effort as the “take action element,” asking how the Army can move from autonomously detecting an adversary to “doing something” through an autonomous response “within prescribed parameters of a human operator, or even in conjunction with that.” That framing highlights procurement needs beyond AI models alone: mission integration, secure testing environments, evaluation methods, auditability, operator controls, and processes for retraining or limiting agent behavior may all be relevant areas for vendors to monitor. The source material does not specify technical requirements, contract vehicles, or solicitation timelines.

Project Griffin also reflects a wider ARCYBER strategy. At the Association of the United States Army conference in Washington, D.C., Lt. Gen. Christopher Eubank said the command is creating an agent workforce that works “hand in hand with the human workforce.” Task Force Lexington began in April with a handful of cyber roles, including exploitation analysts and data engineers, and expanded toward mapping an entire Cyber Protection Team, with roles such as analytic support and network analysis. Officials have described agents working in sandbox environments, with human teams checking their work daily and retraining them as needed. Eubank said the command is still learning where autonomy is appropriate, and that “all work right now is safeguarded human-in-the-loop.” These statements reinforce that oversight, incremental deployment, and evidence of performance are central to the Army’s approach—and likely important considerations for companies positioning cyber automation capabilities.

• Project Griffin is a pilot for AI agents designed to detect and respond to cyber threats within human-defined operating limits.
• The Army reportedly chose two companies from a final field of 10, after receiving more than 100 industry submissions; company names and award values are not provided.
• The planned October 16, 2026 announcement date has passed, but the reporting supplied does not confirm that awards were announced.
• Agents are expected to undergo evaluation and sandbox testing with human oversight before any use with Army network data; immediate operation on government networks is not established.
• Brandon Pugh characterized the planned award as the “take action element,” emphasizing autonomous response within operator-prescribed parameters or in coordination with an operator.
• ARCYBER’s broader agent workforce effort is associated with Task Force Lexington, launched in April, and has expanded from selected roles toward an AI-enabled Cyber Protection Team.
• Lt. Gen. Christopher Eubank described daily human checks, retraining, and incremental decisions about which tasks agents may perform independently.
• Follow-on areas identified in the source signal include cyber deception and agentic red teaming; one signal says funding is aligned, but no amount or award mechanism is specified.
• Contractors can track future notices and assess capabilities in cyber automation, sandbox testing, performance evaluation, human oversight, and agent retraining without treating the reported selections as verified named awards.