- US intelligence agencies and the military are developing AI programs.
- But they need specialized, secure systems to master sometimes chaotic AI models.
- Microsoft created an AI system cut off from the Internet for American intelligence.
The U.S. military and intelligence services are eager to harness the potential of AI, and companies are developing new technologies to make it happen.
While many industries can freely experiment with AI and use public tools, the high stakes and sensitivity of intelligence work and warfare represent a major obstacle.
For companies that can sufficiently protect their data and defend against the well-documented errors and hallucinations of AI models, a significant new market awaits. Its tasks are as varied as sorting through reams of National Security Agency intercepts to detecting terrorist threats and guiding battlefield decisions in real time.
Companies like Microsoft have built siled AI products for the intelligence community, and Palantir has also outlined its ambitions. Similar efforts created an uproar within Google years ago.
An emerging company
This month, Radha Plumb, a senior Pentagon official specializing in AI, highlighted the low amount of classified computing power as an obstacle as the Pentagon prepared to conduct further testing. Defense One reported; Plumb has since resigned.
As demand from defense and intelligence agencies increases, business opportunities are also expected to increase.
Officials hope AI can supercharge tasks ranging from analyzing secret data to battlefield targeting, an approach used by the Israeli Defense Forces in their devastating air war against Hamas-ruled Gaza.
“The United States is considering integrating AI into a wide range of national security tasks,” said Ian Reynolds, a postdoctoral researcher in the Futures Lab at the Center for Strategic and International Studies.
He said the Pentagon has about 800 AI-related projects underway and is rolling out uses of the technology identified in a 2023 test program called Project Lima.
“There are indications that the technology is operational in some circumstances even today,” Reynolds said.
Defense One reported that the US military was trying to understand how AI could help its leaders make decisions more quickly in a potential conflict with China, with tests in the Pacific region.
“The idea is to speed up the decision-making process and gain what the DoD calls a ‘decision advantage,’ or the ability to make better decisions faster,” Reynolds said.
One of the Pentagon’s main goals is to improve the flow of information within different parts of the military.
Not only the United States, but also countries like China and the Gulf states, are fighting to dominate the new technology and experiment with how it can be used by spies and the military.
Reynolds said one of the primary functions would be to analyze quantities of classified data.
“I think the goal here is to get the most critical data, information or broader patterns of data, at a faster pace than an analyst,” he said.
Power and danger
The dangers, however, are numerous and serious: classified data could accidentally drift into unclassified uses for an AI. It could leak or be stolen.
AI models could also exhibit biases that are difficult for humans to detect or misunderstand the nuances of communication reports, thereby distorting the decision-making process.
“We are not entirely sure the extent to which human decision-makers can be nudged toward certain decision paths by AI-based decision support systems,” Reynolds said.
And the secrecy of the programs being deployed is another concern among critics.
Amos Toh, senior counsel for the Liberty and National Security Program at the Brennan Center for Justice, told Business Insider that “what little we know about military uses of commercial AI indicates a real risk of exposing classified information to adversaries.
“The use of AI in intelligence analysis can also scan large amounts of personal and sensitive data while amplifying discriminatory predictions about who poses a threat to national security,” he added.
Microsoft announced in December that it had created a solution: a siled AI capable of managing classified data securely.
He said it was the first time in the world that a major AI model had operated entirely separate from the Internet, signaling the start of a new type of spy-friendly AI.