Military Pushes AI Autonomy, Sidestepping Procurement Rules

Defense organizations are accelerating autonomous-systems acquisitions, while intelligence agencies use procurement authorities to sidestep some federal contracting requirements. The push puts military users, contractors and government agencies in a faster-moving system where software, data and machine-made analysis must be brought together—and checked.

The apparatus wants speed

Stephanie Kostro, president of the Professional Services Council, told Federal News Network that Government Accountability Office reports show major systems can take more than a decade to move from concept to fielding. She said that timeline is too long for autonomy. Contractors and military end users are now discussing how quickly they can understand, acquire and field capabilities.

Autonomy is spreading across undersea, surface, land, air and space domains, Kostro said, moving beyond the earlier focus on drones and autonomous flying vehicles. Secretary of War Hegseth announced an Autonomous Warfare Command, and the Army now has a component command dedicated to autonomy and robotics. The chain of command is building dedicated structures as the technology expands.

Development projects bring commercial technology developers into direct comparison with frontline military units. Kostro described a move away from traditional hardware procurement and toward software-driven mission services. Contractors are shifting from selling distinct physical assets to subscription-based models built around continuous software integration, which she called “mission as a surface” and also referred to as “X as a service.”

A commercial-first approach, she said, could make technologies as commercial as possible before tailoring them to military requirements. But some military needs, including lethality, aren't generally available commercially. That procurement process is where agencies and contractors negotiate the boundary between commercial products and military uses.

Faster buying, with a verification bill

The intelligence community is also changing its contracting and procurement, seeking faster timelines, commercial technology and mission-driven outcomes. Agencies use other transaction authorities to avoid some Federal Acquisition Regulation compliance requirements. The effort connects to executive orders on AI signed by the president over the previous year and a half.

Agencies need to coordinate networks of autonomous AI agents and make them interoperable as they rely on cloud and integrated data environments, Kostro said. She described a “verification tax”: the time agencies spend checking data and analysis received from other agencies. That includes human review of AI outputs for hallucinations, defects and algorithmic bias. Faster acquisition doesn't erase the work of checking what the systems produce.

Kostro said the intelligence community was trying to cut acquisition cycles from a couple of years to six months. Space-based assets and intelligence fusion were areas of progress, she said, with humans checking AI-generated material. The speed target and verification burden remain side by side in the same institutional pipeline.

The Professional Services Council planned to host a conference on AI and autonomy on October 30, with a separate intelligence-related forum planned for January. No grassroots organizing, mutual aid or community response appears in the account; the speakers are agency and contractor participants, and the people described as users are military end users.

Kostro said, “I think part of the contractor community's perspective is really thinking through, what does this do to acquisition timelines?” She also said: “Speed is okay. I think somebody had said long, long time ago, somebody way smarter than all of us said something about like speed kills, but also delay kills. And you wanna fail early and fail small so that you can build towards that success.”