AI is driving a tsunami of private-sector open-source intelligence. Now the federal Intelligence Community just has to figure out how to ride the wave.

Mid-level officials from the NSC and State Department will lead the talks, which follow on Xi-Biden summit last November. No public joint statement is expected, let alone a formal agreement.

“How are we going to reduce civilian deaths to the smallest amount humanly possible?” asked Palantir’s Alex Karp this week. “That is a tech problem.”

The Defense Department wants to explore Large Language Models for everything from paperwork to war plans – without being misled by hallucinations or having sensitive information sucked up by commercial LLMs hungry for training data.

The Defense Information Systems Agency’s five-year plan includes the ambitious goal to build a global network “unconstrained by bandwidth [and] impervious to denial” by hostile forces.

The 1st Light Armored Reconnaissance Company, which normally rides 6-wheeled LAVs, will work with the Marine Corps Warfighting Lab to test an Australian boat designed for high-speed, high-tech scouting and special ops.

“No matter what theater we talk about – PACOM, Europe – everything needs to be smaller, lighter [and] very modular, so…we can throw it in the back of a pickup truck,” said Col. Devin Licklider, program manager for MAGTF Command & Control.

Instead of demanding an exhaustive “AI Bill of Materials.” the Army will only ask contractors for a “baseball card” of key stats on their AI — while building up its in-house capacity to check for bad code or “poisoned” data.

“Any commercial LLM that is out there, that is learning from the internet, is poisoned today,” Jennifer Swanson said, “but our main concern [is] those algorithms that are going to be informing battlefield decisions.”

“It’s great to have internet day to day in peacetime,” said Lt. Gen. Robert Skinner, director of the Defense Information Systems Agency, “but it’s more imperative to have it when bullets are flying.”

MDA Chief Lt. Gen. Heath Collins said more maneuverable missiles and drones have changed the missile defense game: Instead of just preparing to hit “fastballs,” he said, “now we’re hitting sliders and curveballs.”

The service’s new policy empowers “mission area data officers” for warfighting, intelligence, business operations, and enterprise IT, as well as institutionalizing what have been “ad hoc” data duties across the service, David Markowitz told Breaking Defense.

With a max value of $4.1 billion over five to 10 years, the C2BMC-Next contract will upgrade the global missile defense system to tap new satellite feeds, track hypersonic and cruise missiles, and employ AI — all potential building blocks of a future CJADC2 meta-network.

With ever-larger numbers of drones, unmanned submersibles and robot boats, said 4th Fleet Rear Adm. Jim Aiken, “what we didn’t realize was the volume of data we were going to get at headquarters.”

“What we need is the applications — not new computers, that takes longer to install on ships,” said Rear Adm. Doug Small, commander of NAVWAR.