The National Security Agency wants access to artificial intelligence models across the entire commercial market — and it is negotiating with leading developers right now to get it. "We want access to all the models, and we're going to take advantage of that," NSA deputy director Tim Kosiba said Thursday on a panel at an Intelligence and National Security event in Bethesda, Maryland, as reported by Nextgov/FCW.
Kosiba's remarks offer one of the clearest public indications yet that the NSA is actively engaging major AI developers as it assumes a central role in the government's new framework for evaluating advanced models. The agency is implementing new White House directives on AI, and its deputy director did not name which companies are participating in the emerging voluntary model-testing program. Per Nextgov, the comments also follow months of internal deliberations among officials about what role the US intelligence community's primary signals intelligence and cyber organization should play in advancing American AI competitiveness. For more on the policy battles shaping the technology, see our AI industry coverage.
What the June Executive Order Requires
The conversations are happening as the NSA helps implement a sweeping June executive order that directs national security agencies to create classified tests for determining when AI systems have developed sufficiently advanced hacking capabilities to warrant additional scrutiny.
Under the order, a voluntary arrangement lets developers give the federal government access to qualifying models for as many as 30 days before releasing them to other trusted partners. Crucially, the order explicitly prohibits the creation of a mandatory government licensing or approval process for new AI models — a line the administration has drawn to keep the regime cooperative rather than coercive.
The 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.
What "Access" Actually Means
The term may cover several different arrangements, per Nextgov. The 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. Kosiba did not say which models the NSA currently uses or disclose whether any developer has provided the agency with access to an unreleased system.
"The transformative change that we see today is what these models can actually do," Kosiba said, noting that while the agency has used AI in various forms for decades, the rapidly improving capabilities of newer systems are driving a more aggressive effort to obtain and deploy the technology. "There are a lot of conversations happening amongst industry, especially the frontier model companies."
A Broader National Security Push
Kosiba's comments fit into a wider June national security directive that 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. That directive also encouraged the government to avoid dependence on any single company by sourcing technology from a diverse group of suppliers.
Notably, 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 weights available for anyone to download and modify — a category the administration has so far kept out of the classified evaluation pipeline.
Government access to advanced AI is expanding in parallel through ordinary procurement channels as well: OpenAI's latest GPT-5.6 models became available to federal customers this month through the company's FedRAMP-authorized enterprise service, following their public release in July.
Why It Matters
The NSA's posture crystallizes a tension at the heart of US AI policy. On one side is the national security argument: if advanced AI systems can autonomously hack, agencies with cybersecurity mandates want eyes on them before release, and a voluntary 30-day window is the mechanism designed to provide it. On the other side is the innovation argument — one the executive order itself acknowledges by banning a mandatory licensing regime, and which the open-weight exclusion reinforces by leaving the open ecosystem untouched.
What Kosiba's framing makes clear is that the NSA sees comprehensive model access not as an aspiration but as an operating requirement. The agency's decades of institutional experience with AI — and its statutory role in securing classified networks — position it as the natural referee for judging which systems are capable enough to need scrutiny before release. The unanswered questions are the ones that will shape the program's credibility: which developers have signed on, what happens inside the 30-day evaluation windows, and whether the voluntary framework holds as capabilities — and stakes — keep climbing.
---
Stay Ahead of AIGet the latest AI news, analysis, and breakthroughs — all in one place.
Read more AI news →
