It Begins: Pentagon To Give AI Agents A Role In Decision Making, Ops Planning

Updated The American military has signed a deal with Scale AI to give artificial intelligence, as far as we can tell, its most prominent role in the Western defense sector to date – with AI agents to now be used in planning and operations. 

The value of the contract, awarded as part of the US Defense Innovation Unit's Thunderforge project, wasn't specified, though given its considerable scope it's likely to be a large one. According to data labeling and AI training outfit Scale today, the contract will see it leading a team including Palmer Luckey's Anduril and Copilot-obsessed Microsoft to implement the US Department of Defense's "first foray into integrating AI agents in and across military workflows."

According to the DIU's Thunderforge project leader Bryce Goodman, transforming military decisionmaking with AI is something that's pivotal to sustaining US military dominance. 

Thunderforge brings AI-powered analysis and automation to operational and strategic planning

"Today's military planning processes rely on decades-old technology and methodologies, creating a fundamental mismatch between the speed of modern warfare and our ability to respond," Goodman said. "Thunderforge brings AI-powered analysis and automation to operational and strategic planning, allowing decision-makers to operate at the pace required for emerging conflicts."

The end goal of the project, the DIU said, is to help military decision-makers pore over and assess more info more quickly and make judgement calls more rapidly based on AI suggestions. Thunderforge AI will be used to support mission planning and campaign development, help allocate resources at the theater level, and make strategic assessments. We're given the strong impression humans make all the final decisions, albeit ones guided by software.

The AI agents – a fancy word for bot – can carry out table-top war-gaming to simulate outcomes for leaders, plan scenarios, "and refine proposed courses of action," the DIU said. 

It appears the automated system will be tested in the real world, with both the DIU and Scale indicating it's rolling out at some unspecified time to the US Indo-Pacific Command headquartered in Hawaii, and to what for now is the US European Command based in Germany.

The Pentagon has plans to scale it across all 11 of its combatant commands after its initial deployment. 

... a decisive shift toward AI-powered, data-driven warfare, ensuring US forces can anticipate and respond to threats with speed and precision

"Thunderforge marks a decisive shift toward AI-powered, data-driven warfare, ensuring US forces can anticipate and respond to threats with speed and precision," the DIU said. 

Anduril's role in Thunderforge will see it supplying its Lattice software platform, while Azure giant Microsoft will be handling the actual large language models. Scale's role will be in leveraging its agentic applications and generative AI evaluation expertise, the DIU said. 

"Our AI solutions will transform today's military operating process and modernize American defense," said Scale founder and CEO Alexandr Wang. 

Safety not guaranteed?

The use of AI for defense purposes has been a divisive issue in the tech world, with Googlers fired last year for protesting the Chocolate Factory's role in providing cloud services to the Israeli Ministry of Defense, and Microsoft employees terminated last month for making similar complaints.

Indeed, Israel, for one, already uses Made-in-America AI models in war.

Google has walked back previous promises to not use AI for weapons development, spying and violating international norms, as has OpenAI. 

• Maybe cancel that ChatGPT therapy session – it doesn't respond well to tales of trauma
• Flanked by Palantir and AWS, Anthropic's Claude marches into US defense intelligence
• Defense AI models 'a risk to life' alleges spurned tech firm
• Mental toll: Scale AI, Outlier sued by humans paid to steer AI away from our darkest depths
• We can't make this stuff up: Palantir, Anduril form fellowship for AI adventures

The American military has been increasingly integrating artificial intelligence into its operations, with Scale even scoring a Dept of Defense contract last year to work on integrating generative AI for various non-combat purposes, such as compiling after-action reports, measuring performance, and the like. 

As with all things AI related, there's the question of the safety and reliability of LLMs and similar machine-learning models, as well as their ability to return real and correct information, something neither Scale nor the DIU addressed in their statements about Thunderforge. 

Scale did mention that Thunderforge AI will always act "under human oversight," but that's the only mention of safety by either organization. We reached out to both Scale and the DIU for further details, and as yet haven't heard back. ®

Updated to add at 2305 UTC

The Defense Innovation Unit has been in touch to tell us it understands the concerns around AI safety in military uses, telling us it has been a leader in developing safeguards for military AI.

As part of that, the DIU said Thunderforge, even with humans in the loop, is being designed for intelligibility.

... maintaining human oversight is critical

"Accuracy and reliability are core design principles, and maintaining human oversight is critical," a DIU spokesperson told us. "That’s why significant effort is being put into making Thunderforge’s reasoning process legible, so users can trace its logic, assess confidence levels, and make informed decisions."

Thunderforge will also be integrated "with authoritative databases to ensure it operates with up-to-date, validated information across all classification levels," the spokesperson told us.

Ultimately, however, "the Department of Defense still relies on staff planning processes that date back to the Napoleonic era," the DIU said in response to our questions.

"At the operational level, we're seeing the rise of autonomous systems and swarm warfare, where decision cycles are measured in seconds, not days," the spokesperson added. "Thunderforge is designed to bridge this gap by accelerating strategic planning without sacrificing human oversight."

Botnote

The LA Times just pulled offline a bot it was running on its website, which summarized and offered counterpoints to articles, because it defended the Ku Klux Klan under a column about the notorious White supremacist group being run out of Orange County a century ago. The automated editorial AI agent had been live for one day.