- US intelligence agencies and the military are developing artificial intelligence programs.
- But they need specialized, secure systems to rein in sometimes chaotic AI models.
- Microsoft created an artificial intelligence system isolated from the Internet for US intelligence.
The US armed forces and intelligence agencies are eager to harness the potential of artificial intelligence, and companies are developing new technology to enable it.
While many industries can freely experiment with AI and use common tools, the high stakes and sensitivity of intelligence and war work present a major barrier.
For companies that can keep data sufficiently secure and defend against the well-documented errors and hallucinations of AI models, a large new market awaits. Its missions range from sorting through NSA intercept collections for terrorist threats to guiding decisions on the battlefield in real time.
Companies like Microsoft have built closed AI products for the intelligence community, and Palantir has also bet on its ambitions. Similar efforts have caused uproar for years within Google.
Emerging businesses
This month, Radha Plumb, a senior Pentagon official focused on artificial intelligence, pointed to the small amount of classified computing power as an obstacle as the Pentagon prepares to conduct new tests. You mentioned one defense; Plumb has since stepped down.
As demand from defense and intelligence agencies increases, so should business opportunities.
Officials hope AI can augment tasks from analyzing troves of classified data to battlefield targeting, an approach the Israel Defense Forces used in its devastating air war on Hamas-led Gaza.
“The United States plans to integrate artificial intelligence into a wide range of national security-related tasks,” said Ian Reynolds, a postdoctoral fellow in the Futures Lab at the Center for Strategic and International Studies.
He said the Pentagon had about 800 AI-related projects underway, and was deploying uses of the technology identified in a 2023 test program called Project Lima.
“There are some indications that the technology works in some circumstances even today,” Reynolds said.
Defense One reported that the US military was trying to figure out how artificial intelligence could help its leaders make decisions faster in a potential conflict with China through tests in the Pacific region.
“The idea is to speed up the decision-making process and achieve what the Department of Defense calls ‘decision advantage,’ or the ability to make faster and better decisions,” 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 countries including China and the Gulf states are racing to control the new technology and experiment with how it can be used by spies and the military.
One of the primary jobs will be analyzing sets of confidential data, Reynolds said.
“I think the goal here is to get the most important data and information or broader patterns across the data, at a faster rate than the analyst,” he said.
Power and danger
However, the risks are many and severe – confidential data may inadvertently drift into unclassified uses of AI. It could leak or be stolen.
AI models can also show bias in ways that are difficult for humans to understand or may misunderstand the nuances of communications reports, distorting the decision-making process.
“We’re not entirely sure the degree to which human decision makers might be nudged toward certain decision paths by AI-powered decision support systems,” Reynolds said.
The secrecy of the programs being offered is another concern for critics.
“What little we know about military uses of commercial AI suggests a real risk of exposing classified information to adversaries,” Amos Toh, a senior adviser in the Freedom and National Security Program at the Brennan Center for Justice, told Business Insider.
“The use of AI in intelligence analysis may also lead to the collection of vast amounts of personal and sensitive data while amplifying discriminatory predictions about who poses a threat to national security,” he added.
Microsoft said in December that it had come up with a solution: walled artificial intelligence that can handle confidential data securely.
It said it was the first time in the world that a major model of artificial intelligence had been run separately from the Internet, signaling the beginning of a new type of spy-friendly AI.
