Pentagon AI rollout races ahead of oversight deadlines
The Pentagon’s artificial-intelligence rollout is moving ahead of the first major oversight deadlines in President Donald Trump’s June 5 national-security memorandum, with new classified-network agreements and broad internal use of AI already announced.
The unresolved question is whether the controls required by that memorandum are advancing at the same pace.
National Security Presidential Memorandum 11 directs the Department of War and intelligence agencies to accelerate national-security AI adoption while producing new rules for autonomous weapons, governance, procurement, security, training and testing. As of August 5, 2026, the public record reviewed does not confirm that the memorandum’s first required policy updates have been completed.
What the memorandum requires
NSPM-11 gives the Secretary of War 90 days to issue an update to Department of Defense Directive 3000.09, the directive governing autonomy in weapon systems. That makes September 3, 2026, the first major checkpoint.
The memorandum also requires the Committee on National Security Systems and the Office of Management and Budget, working with other national-security officials, to issue an AI governance policy and a classified annex within the same 90-day period. Within 120 days, it calls for additional work on procurement, access to computing capacity, industry security, data exchanges, training, risk management and testing. Those deadlines fall on October 3, 2026.
These are requirements and deadlines, not evidence that the resulting policies are already in force. The White House memorandum does not by itself establish that the Department of War or other agencies have completed the required documents.
Deployment is already expanding
In a May 1 announcement, the Department of War said it had reached agreements with eight companies—SpaceX, OpenAI, Google, NVIDIA, Reflection, Microsoft, Amazon Web Services and Oracle—to deploy artificial-intelligence capabilities on classified networks.
The department also said more than 1.3 million personnel had used GenAI.mil during its first five months, generating tens of millions of prompts and deploying hundreds of thousands of AI agents.
A separate AI Acceleration Strategy describes three major efforts: Agent Network for AI-enabled battle management and decision support, GenAI.mil for department-wide access to frontier models, and Enterprise Agents for secure workflow deployment. The strategy also includes artificial-intelligence infrastructure and workforce initiatives.
The department’s announcements document expanding access and experimentation. They do not, by themselves, show that the programs satisfy NSPM-11’s procurement, governance or testing requirements.
The company count has differed across accounts. The Department of War’s May 1 release lists eight companies, while an Associated Press report described a separate seven-company arrangement and noted that Anthropic was absent from the department’s list. The accounts should not be treated as identical without additional documentation.
Why Directive 3000.09 matters
Directive 3000.09 is central to the debate because it addresses how autonomous and semi-autonomous weapon systems are developed, tested and used. The safeguards discussed in the directive and related oversight include human judgment, geographic limits, termination procedures, reliability and robustness testing, and protection against adversarial manipulation.
The pending update has drawn congressional scrutiny. In a June 12 letter, Sen. Ruben Gallego questioned whether a compressed rewrite could weaken human judgment, shorten testing, or increase risks involving friendly fire and civilian harm. He also asked about coordination with allies and protections for civilians in host nations.
Gallego’s letter is an oversight document and a set of questions, not a finding that wrongdoing has occurred. The available records also do not establish that autonomous weapons are currently operating without human control or that a specific deployment failure has occurred.
Inspectors are reviewing open recommendations
On June 29, the Department of Defense Office of Inspector General, using the secondary title Department of War Inspector General, opened a follow-up evaluation of open recommendations concerning AI policy, governance, acquisition and implementation procedures.
The memorandum says the review will examine the status of recommendations from earlier evaluations, including work involving AI governance and acquisition. It also states that the use of the Department of War title does not change the Department of Defense’s statutory name, authorities or the inspector general’s responsibilities.
The evaluation is an active oversight action, not a final compliance finding. Its eventual conclusions could help clarify whether existing controls, acquisition practices and governance arrangements match the department’s expanding AI use.
What remains unknown
As of August 5, the public sources reviewed do not verify publication of the revised Directive 3000.09, the national-security AI governance policy, or the classified annex required by NSPM-11. Public records also do not establish whether the department’s announced programs have completed the memorandum’s related testing, risk-management, procurement and reporting steps.
Many details may remain classified, but the existence, status or public confirmation of the required policy documents would still be an important accountability marker.
What to watch next
September 3 will be the first visible test of whether deployment and oversight are moving together. Readers should watch for confirmation of the revised Directive 3000.09, the national-security AI governance policy, the status of its classified annex, and provisions governing human oversight, testing, geographic limits, termination and system robustness.
October 3 is the next major checkpoint for procurement, industry-security, data-exchange, training, risk-management and testing deliverables.
The issue reaches beyond classified facilities. Military AI decisions can affect servicemembers, allied forces, civilians in conflict zones, contractors, technology companies and taxpayers. The immediate public question is not whether the government is using AI. Official records show that it is. The accountability question is what documented safeguards and independent reviews accompany that use.
Sources
- National Security Presidential Memorandum/NSPM-11 — The White House
- Classified Networks AI Agreements — U.S. Department of War
- Follow-up Evaluation on Artificial Intelligence Recommendations — Inspector General
- U.S. military and 7 companies make deals to use AI in classified systems — Associated Press
Discover more from Interactive News
Subscribe to get the latest posts sent to your email.
