Juelich’s findings suggest extremist organizations have already moved beyond simply experimenting with AI. Instead, she found the technology becoming embedded in routine operations as a general-purpose tool to solve problems, improve efficiency, and overcome logistical obstacles.

“What surprised me the most was that AI adoption and use has been faster, more extensive, and more systematic than I thought,” Juelich told the Washington Examiner. “AI misuse by terrorist groups is no longer some kind of hypothetical future threat, but a present and growing reality.”

Much of the public discussion around AI and terrorism has focused on catastrophic scenarios, such as terrorists using AI to develop chemical or biological weapons. Juelich argued that this emphasis overlooks how extremist groups are actually adopting the technology.

“There was very little evidence on what was happening on the ground,” she said. Existing research, she argued, has largely focused on what AI models are theoretically capable of rather than how terrorist organizations are using them in practice.

Instead, she found AI functioning as an “operational assistant,” helping militants answer everyday questions, troubleshoot problems, and improve existing capabilities.

“They really understood that it’s a general-purpose technology that they can use for any kind of questions that come up,” she said.

That “operational middle ground” is where Juelich believes governments and AI companies should concentrate their attention. Juelich said her interviews demonstrate the need to better understand how AI is already helping extremist groups overcome routine operational challenges before those capabilities evolve into something more dangerous.

An intelligence problem with no single solution

The research also underscores a high-level challenge: No single government agency, AI company, or research institution has a complete picture of how terrorist organizations are using artificial intelligence.

Rebecca Hersman, a national security expert who has proposed creating an AI threat fusion center, said critical information is scattered across intelligence agencies, technology companies, and independent researchers.

“Critical information sits in very different locations,” Hersman told the Washington Examiner. She added that AI companies may detect suspicious user behavior, intelligence agencies may possess classified reporting, and researchers may uncover insights through interviews like Juelich’s research.

“What we don’t have is a way to pull that together … in a way that’s safe and secure,” she added.

Juelich’s interviews, Hersman said, illustrate the type of information that is difficult for any one organization to collect but becomes far more valuable when combined with other intelligence.

“Sitting alone, it’s not perfect,” Hersman said. “But put together more holistically, knitted together with other information, it could become incredibly valuable.”

To bridge those gaps, Hersman has proposed creating a fusion center where cleared government officials and designated representatives from AI companies could share threat information while protecting classified intelligence and proprietary data.

Such a center, she said, could produce “threat-informed risk assessments” of how extremist groups are using — or may seek to use — AI, while also conducting classified model evaluations and reducing the risk of publicly disclosing sensitive information that could inadvertently aid adversaries.

Hersman said Juelich’s report reinforced her own research showing that terrorist organizations are using AI less for futuristic weapons programs than for practical operational support.

“They’re really using it for what I would call operational assistance,” Hersman said, referencing Juelich’s report and pointing to tasks such as surveillance, target selection, and basic bomb-making guidance. While those activities may not be technologically sophisticated, she said, AI can remove obstacles that have traditionally complicated terrorist operations.

That, she argued, makes it essential for governments and AI companies to look for those patterns now rather than waiting for more advanced forms of misuse to emerge.

“We need to lean forward and try to get ahead of where these actors might go before they actually go there,” Hersman said. “We don’t want to be looking at that problem in hindsight.”

How AI companies are responding

The findings come as AI developers face pressure to demonstrate that their systems cannot be exploited by terrorist organizations.

OpenAI told the Washington Examiner its policies prohibit terrorist and violent extremist activity and that its systems are designed to refuse requests related to building weapons or facilitating violence. The company said it can disable accounts, block related accounts, and work with law enforcement when it detects misuse.

Anthropic said it relies on automated safeguards, dedicated threat intelligence teams, and outside partnerships to identify and disrupt dangerous activity.

“Our safeguards are built to refuse dangerous requests, including those tied to violence, attack planning and building explosives,” an Anthropic spokesperson told the Washington Examiner.

The company also emphasized that defending against AI-enabled threats requires cooperation across the industry.

“No single company can counter these threats alone,” the spokesperson said.

The Cybersecurity and Infrastructure Security Agency told the Washington Examiner it is also working with government and private-sector partners to monitor emerging AI developments and assess how the technology can strengthen national cyber defenses against malicious actors.

The race to stay ahead

Juelich warned that AI capabilities are advancing more quickly than traditional methods of assessing terrorist threats.

“It’s great if one provider has really strong safeguards,” she said. “But it doesn’t make us safer as a society if there are still other companies who don’t really care about safety. We are only as strong as the weakest one.”

Hersman agreed that strong guardrails at AI companies are important, but argued they are only one piece of the solution. To stay ahead of emerging threats, she said, governments and industry need a far better understanding of how specific terrorist organizations are already using AI to overcome operational barriers.

HOW THE BIG BEAUTIFUL BILL BECAME LAW: READ MORE FROM THE WASHINGTON EXAMINER’S 10-PART SERIES

She said creating a fusion center would require combining classified intelligence, company data, and outside research in carefully controlled settings before extremist groups become sophisticated enough to pursue higher-risk capabilities.

“We don’t want to be doing that or looking at that problem in hindsight,” Hersman said, adding that “the companies are doing great work when it comes to setting really firm guardrails and trying to prohibit access to that information, but that is the difficult process, especially when we’re doing it generically.”