Elodie Collins 
- Proponents say agentic AI can improve IC operations and security
- Adversaries are also using agentic AI to exploit vulnerabilities in U.S. systems
- Potomac Officers Club will host multiple agentic AI-focused panels at the 2026 Intel Summit
The rise of agentic artificial intelligence systems is disrupting the modern security landscape. The technology, capable of autonomous decision-making, is thought to offer the intelligence community profound benefits, strengthening its capability to defend national security and advance U.S. interests globally.
This article examines three key use cases of agentic AI within the IC, from enhancing agency efficiency and management of open-source intelligence to bolstering the nation’s cyber infrastructure.
The 2026 Intel Summit on Sept. 24 will feature multiple agentic AI-focused panels exploring the technology’s growing impact on the IC. From scaling the deployment of agentic AI to achieve a decision advantage to mitigating AI-enabled cybersecurity threats in classified environments, the summit will give attendees insights into how industry can support intelligence agencies. Tickets are now available here!
What IC Leaders Have Said About AI
CIA Director and Wash100 Award winner John Ratcliffe highlighted the critical need for the agency to adopt AI and other emerging technologies, NextGov/FCW reported.
“We simply can’t afford to wait for a risk-free approach when it comes to emerging technologies. It doesn’t exist,” he said at the Amazon Web Services Summit in June. “We have to move fast. We have to be aggressive.”
Ratcliffe noted that the agency will “take smart risks, experiment and course-correct” as it embeds AI into its operations.
The IC has been utilizing AI for years. The 2019 Operation Sable Spear, which exposed illicit Chinese fentanyl trafficking, used AI to parse through non-classified data and identify companies and people engaged in illegal activities, according to PBS.
“You wouldn’t be able to do that without artificial intelligence,” Brian Drake, Defense Intelligence Agency’s then-director of AI, noted.
The CIA-made generative AI platform Osiris is in use across the 18 U.S. intelligence agencies. Jennifer Ewbank, former CIA deputy director for digital innovation and a two-time Wash100 winner, discussed in a 2025 white paper how Osiris processes large troves of open-source intelligence, or OSINT, and provides summaries to support decision-making.
Ewbank spoke about the impact of AI in the security industry at POC’s 2025 Cyber Summit. When she was still with the CIA, she delivered a keynote address at the 2023 Intel Summit. Intelligence leaders of the same caliber will be present at the 2026 Intel Summit on Sept. 24. Officials from the National Security Agency, National Reconnaissance Office and other agencies will share insights about the present threat landscape, and identify opportunities where industry can deliver next-generation capabilities. Click here to register!
DIA has also deployed a large language model dubbed ChatDIA into the Joint Worldwide Intelligence Communication System, a top-secret government network, reported Federal News Network.
Efforts to implement AI within the IC align with a June memo from the White House that tasked defense and intelligence agencies to accelerate the technology’s adoption to strengthen national security.
How the CIA Plans to Deploy AI Coworkers
The CIA plans to integrate AI into analyst workflows in the coming years, according to CIA Deputy Director Michael Ellis. At an event in April, the top intelligence official shared that AI-powered coworkers will support agency personnel with basic tasks, such as drafting and editing. The intelligent tools, he added, will be housed in the agency’s analytics platform, but they will not replace human analysts.
Ellis emphasized that AI will not “do the thinking for our analyst.”
The CIA had more than 300 AI projects in 2025, the official revealed. Within 10 years, he said the agency will adopt AI as “an autonomous mission partner,” and officers will be tasked with managing AI agent teams in a hybrid workforce model designed to speed up and scale intelligence.
Can Agentic AI Improve Intelligence Data Management?
Proponents say that Agentic AI can improve analyst workloads and accelerate the process of collecting, analyzing and responding to threats.
In a company blog post, Everforth leaders described AI as the “ultimate analyst” capable of processing massive and diverse datasets, including satellite imagery, financial transactions, social media posts and intercepted communications.
OSINT analysis is also one of the three high-value use cases for agentic AI that Bill Pessin, senior vice president of Salesforce National Security, identified in an article published by NextGov/FCW. The executive explained that AI can rapidly sift through data from multiple sources, identify anomalous patterns, and generate comprehensive and timely intelligence reports. He added that the technology will enable human analysts to focus on making strategic decisions.
The DIA is already exploring AI to better manage and leverage OSINT through its Digital Modernization Accelerator launched in March.
“We talked about the ocean of data, so how do we sift through tens of thousands of pieces of data or tens of millions?” Eric Miller, senior adviser for OSINT at DIA, said. “We’re trying to leverage AI to help make sense of that. Once we do that, then we now have to get that report out, and that’s another automated process that we need to do, because if we’re creating more products, we need to get it out even quicker.”
How AI Can Fight AI-Powered Cyberattacks
Adversarial states frequently use cyberattacks to infiltrate an organization. Data from the open-source threat intelligence website Hackmageddon revealed that cyber espionage accounted for one in five confirmed cyber incidents in the first half of 2026. In the World Economic Forum’s Global Cybersecurity Outlook 2026, the majority of organizations polled see geopolitics as the top factor influencing their overall cyber risk mitigation strategies.
Agentic AI can thwart cyberattacks, even those carried out by autonomous tools, according to a Congressional Research Service report published in July. CRS explained that agentic AI can detect anomalies and take immediate counteractions autonomously, preventing an attacker from causing lasting damage. Moreover, machine learning can learn from cybersecurity datasets, anticipate attacks and recommend preventive measures.
Mythos, which Anthropic said is its most powerful AI model yet, is aiding organizations across sectors globally to strengthen cybersecurity. Through the joint industry initiative Project Glasswing, Anthropic gives partner organizations access to Mythos to identify zero-day vulnerabilities.
TechCrunch reported that Project Glasswing’s list of partner organizations cover power, water, healthcare and communications industries. Anthropic also provided the U.S. government access to Mythos as part of the program in early April.
What Are the Dangers of Agentic AI?
The proliferation of agentic AI means adversaries are also using the technology against the United States.
In 2025, Anthropic disclosed that threat actors linked to China used its agentic AI assistant, Claude Code, to carry out a major cyber espionage campaign targeting 30 organizations, including government agencies. According to the company, the threat actors bypassed Claude’s security protocols by breaking malicious actions into seemingly innocuous tasks, making the AI believe that it was conducting legitimate cyber operations.
Throughout the campaign, Claude executed thousands of requests.
“And then [this operation] in September, I think what we’re seeing now in this case is to me the most autonomous misuse we’ve seen,” Jacob Klein, head of threat intel and AI safeguards at Anthropic, told CyberScoop.
Edacheril Mathew, chief information officer at the DIA, recently cautioned about cyberthreats associated with agentic AI. Pointing to recent incidents where sandboxed AI broke out of containment and caused cyber incidents, the CIO warned that future cyberattacks may not be conducted by humans but by AI agents.
More about the threat agentic AI and other emerging technologies, such as quantum computing, to the IC at the 2026 Intel Summit on Sept. 24. The summit’s Agentic AI and Quantum Risk: The Blast Radius Problem in Classified Environments panel will feature government and industry leaders to explore strategies for mitigating security risks and safeguarding sensitive networks against threats. Get your tickets today!
