Civilian and government officials are accustomed to using their own phones and tend to rely on commercial messaging apps, even when they shouldn’t. For example, U.S. operators in Poland and elsewhere have communicated with Ukrainian counterparts over the popular communication app Signal, because of its wide adoption and availability on regular consumer phones.

But those conversations—particularly those involving U.S. special operators—may occur over networks that the U.S. military doesn’t own, so-called dirty networks.  This is because they often conduct operations with allied militaries that don’t have access to U.S. top-secret communications platforms or equipment, a challenge that participants at a recent irregular warfare symposium here highlighted across multiple discussions.

“Government executives, they tend to be using their personal devices. They're on Signal,” Eftychis Gregos Mourginakis, co-founder of secure messaging app Kibu, told Defense One during the Global SOF Indo-Pacific Irregular Warfare Symposium.

But there’s no guarantee that the individual on the other end of the phone is the right person, or even a person at all, he said. “How [do State Department officials] trust that when they're talking to their British compatriot, that it’s actually that person? Like, how do you fill that gap? You can't do it with voice. You can't do it on video.”

Recent incidents involving AI agents posing as humans have highlighted the  growing trend—one that a current military official who tracks adversary disinformation campaigns said is a big problem, especially for the military and intelligence communities. 

“AI has made deception scalable. We’re no longer just fighting phishing emails, we're fighting synthetic voices, cloned faces, and entire fake identities convincing enough to fool the people who know the real person best,” the official said.

A small number of U.S. special operators are now testing Kibu, which uses encryption and other verification features to make sure that person—or AI agent—you’re talking to is who they say they are. That includes not just device-to-device encryption, but also biometric authentication and, for military members, can include other forms of identification as well. Organizations can add their own, such as the CAC card requirement. The app even has special passcodes in case the user has their own phone or device but is in a situation where someone else is trying to access it against their will.

The app isn’t tied to a device SIM card, phone number, or even PIN number. And different organizations can host it in different cloud environments. “So if an organization basically signs an agreement with us, all of their end users on this platform are totally air-gapped,” Mourginakis said.

And even as the military faces the challenge of keeping adversary AI agents out of sensitive conversations, they also want to use more AI under what the military calls “meaningful human control.” 

Mourginakis said the app allows individuals to build their own agents within the environment to communicate with unmanned weapons but still report back to the human operator. 

“The identity can look like a human identity, but it can be sandboxed by the human users in that space. They need to biometrically prove it,” he said. “You can see we have a decision quorum. So if I need in this group to authorize certain things, say a certain number of these people need to approve these actions, here they can communicate about it, they can talk about it, they can share the intelligence they need, and then go, hey, yeah, I'm good to do this. Commit my decision. Thumbs up. It's approved.”

While military members face the most immediate need for that sort of verification, Mourginakis said, secure communication is a growing priority for civilians as well.

He sees a shift away from tech products that monetize consumers’ data and push sharing of content toward companies that prioritize privacy.

“Now because of that and the rise of AI slop, the biggest premium service is going to be privacy, trust, security as a service.”

“Some Space Force officials told GAO about challenges they have faced related to purchasing and using commercial space data such as licensing costs, perceived use restrictions, and concerns over long-term access to the data,” the report said. “These challenges led to hesitation to purchase and use commercial data among these potential users.”

The Space Force’s Joint Commercial Operations Cell, the group that buys the majority of the service’s data, spent nearly $77 million between 2023 and 2025 on commercial space services through the Global Data Marketplace, an e-commerce site used by the military for unclassified images. The JCO has a working group to discuss purchases being made by the U.S. government, but Space Force officials aren’t included in that group, according to the report. That’s led to the hesitations and confusion about purchasing intelligence on the platform, the GAO said.

“This working group meets monthly and coordinates purchases. GAO found that many of the challenges reported by Space Force officials could have been addressed with discussions at the working group, but the Space Force officials were not part of the working group and information about the working group is not widely available,” the watchdog report said. 

The GAO recommended that Air Force Secretary Troy Meink, who formerly served as principal deputy director of the National Reconnaissance Office, increase communication and awareness of the working group. The Defense Department concurred with the watchdog’s findings in a June 18 response.

“Increased communication about the working group could lead to additional Space Force and other officials engaging in discussions that could address perceived issues and improve the use of commercial space data and services for national security purposes,” the GAO report said.

The Space Force will need to rely on those commercial assets for future fights, according to the GAO. As adversary counterspace capabilities grow, the Defense Department needs private satellite imagery to fill in blind spots and decrease risk to its own space assets. 

“Threats to DoD’s space assets from strategic competitors have grown substantially in number and intensity,” the GAO report said. “DoD is working to integrate commercial space solutions, such as data and services, to mitigate these threats as well as enhance its capabilities.”

Jon Ludwigson, GAO’s contracting and national security acquisitions director, said in a phone interview on Thursday afternoon that there are positives to using those commercial systems in tandem with the military’s highly-classified spy satellites.

“We're not going to use commercial data for everything, but if we can use it for some things, it reduces the pressure on the exquisite systems to do things,” Ludwigson said. “It also gives us the opportunity to cue others that we're watching in ways that don't require us to disclose that we're using exquisite systems.”

The National Reconnaissance Office, an intelligence agency which builds and launches spy satellites, purchased commercial electro-optical imagery from Vantor, Planet Labs, and BlackSky through its $4 billion Electro-Optical Commercial Layer contract awarded in 2022, according to GAO.

Images purchased by NRO in 2024 showed “changes in Chinese aircraft carriers and other ships’ locations over time, and identify the types of ships” at Sanya Naval Base in Hainan province.

That same year, NRO purchased images that “helped the U.S. Southern Command identify possible clandestine airstrips in Ecuador.” The U.S. government shared the intelligence, which led the Ecuadorian government “to further investigate and disable the airstrips that were likely used by narcotraffickers,” GAO reported.

Identifying those sites and sharing their locations with Ecuador is an example of how governments can benefit from the unclassified imagery, Ludwigson said. It also protects and frees up use of the military’s most crucial spy satellites.

“When we go to share stuff that's coming off an exquisite system, that image is, itself, classified,” he said. “So, if we can use commercial imagery for those conversations with the Ecuadorian government or other governments to say, ‘Hey, look at this. This is what we think it is, and this is what we are concerned about,’ it’'s a commercial asset, then we haven't given up anything about our exquisite assets.”

“We want access to all the models, and we’re going to take advantage of that,” Tim Kosiba said on a panel at an Intelligence and National Security event in Bethesda, Maryland.

Kosiba said NSA has used AI in various forms for decades, but the rapidly improving capabilities of newer systems are driving a more aggressive effort to obtain and deploy the fast-evolving technology.

“The transformative change that we see today is what these models can actually do,” he said.

The remarks offer one of the clearest public indications that NSA is actively engaging major AI developers as it assumes a central role in the government’s new framework for evaluating advanced models. They also follow months of internal deliberations among officials regarding the role the U.S. intelligence community’s primary signals intelligence and cyber organization should play in advancing U.S. AI competitiveness.

“There are a lot of conversations happening amongst industry, especially the frontier model companies,” Kosiba said. He didn’t identify the companies involved, say which models NSA currently uses or disclose whether any developer has provided the agency with access to an unreleased system.

The term “access” may cover several different arrangements. NSA could use a model through a company-controlled service, deploy a version inside a government environment or receive access for security testing of classified networks. 

The conversations are occurring as NSA helps implement a sweeping June executive order directing national security agencies to create classified tests for determining when AI systems have developed sufficiently advanced hacking capabilities to warrant additional scrutiny.

The order calls for a voluntary arrangement through which developers can give the federal government access to qualifying models for as many as 30 days before releasing them to other trusted partners. It explicitly prohibits the creation of a mandatory government licensing or approval process for new AI models.

NSA’s director is responsible for deciding which systems meet the threshold for a “covered frontier model,” in consultation with the Office of the National Cyber Director, the Cybersecurity and Infrastructure Security Agency and other government offices.

A related national security directive issued in June instructed intelligence and defense agencies to form “deep, proactive partnerships” with industry and make the most advanced models available to national security personnel without delay. It also encouraged the government to avoid becoming dependent on a single company by obtaining technology from a diverse group of suppliers.

Kosiba’s comments didn’t establish that the voluntary pre-release program is fully operational or that companies have begun providing models through it. The White House told major developers earlier this month that open-weight models would not be included in the testing program. Unlike closed systems controlled by companies like OpenAI, Anthropic and Google, open-weight models make their underlying settings available for others to download and modify.

Government access to advanced AI has expanded. OpenAI’s latest GPT-5.6 models became available to federal customers this month through its FedRAMP-authorized enterprise service after their public release in July.

The push for broad access comes as the Defense Department remains locked in a legal and policy dispute with Anthropic over restrictions the company placed on certain military uses of its models. The Pentagon designated Anthropic a supply-chain risk after the company declined to relax some of those safeguards, prompting a lawsuit in which a federal judge temporarily blocked parts of the government’s action.

The NSA has gotten carveouts with the company, and is already using versions of Anthropic’s advanced Mythos model on an experimental basis, according to reporting from The New York Times earlier this month. Current and former officials told the newspaper that the agency is using the technology to test defenses protecting U.S. military networks and evaluate its ability to identify weaknesses in networks belonging to China, Russia, North Korea and Iran.

Parts of NSA temporarily lost access to Anthropic’s Mythos 5 amid an export control dispute between the company and the Trump administration, demonstrating how government access to commercially controlled systems can be affected by policy decisions outside the agency. The administration has subsequently lifted the restrictions that had affected access.

AI-enabled cyber risks have become an increasing flashpoint in recent months after Anthropic unveiled its first version of Mythos, an advanced model it withheld from public release because of concerns that its ability to find and exploit previously unknown software vulnerabilities could be misused. 

Since then, Anthropic, OpenAI and other developers have introduced a growing family of models with similar capabilities, forcing policymakers to consider how to test the systems, control access to them and harness their abilities for cyberdefense without giving attackers the same advantage.

As NSA pushes to adopt AI faster, Kosiba said humans will remain responsible for checking that its use complies with surveillance laws and the agency’s internal rules.

“What the models are putting out, how we’re leveraging technology today, will always require that human to double-, triple-check and make sure we are in complete compliance with the law,” he said.

Editor’s Note: This story was updated to add additional context from New York Times reporting.