The military gets AI agents built directly into its systems

Salesforce has announced that its AI suite Agentforce 360 has received a security authorization called Impact Level 5 (IL5). The suite is now built into the company's Missionforce National Security platform, which the US Department of Defense (which now calls itself the Department of War) can use.

IL5 is a US security standard for cloud services. It applies to data that is sensitive but not classified. The authorization means authorities can adopt AI agents without starting over with the security assessments.

What is an AI agent?

An AI agent is a program that does not just answer questions but carries out tasks itself. It can, for example, search through large amounts of data and kick off a process. When the agents are "embedded" in the systems people already use, employees do not need to learn an entirely new tool.

Four examples from Salesforce

The company names four areas where the agents will be used:

- Logistics: Agents can find spare parts shortages on their own and trigger purchases. - Support for soldiers and families: A digital helper available around the clock can answer questions about relocation, housing, childcare and health. - Recruitment: The agents can speed up onboarding and flag bottlenecks in security clearances. - Leadership overview: Commanders can ask questions in plain language across separate data systems and get recommendations in real time.

As background, Salesforce refers to a report from the US watchdog GAO. It found that 74 percent of US Air Force aircraft were not ready for maintenance on time, partly due to spare parts shortages and separate data systems.

Salesforce's head of Missionforce and Government Cloud, Kendall Collins, says: "Trust is the foundation of every successful mission." He believes the authorization allows authorities to use AI without compromising on security.

Caveats

This is a press release from the vendor, so the claims about benefits are the company's own. Slack, which is also part of Salesforce, is moreover only authorized at the lower IL4 level.

Other sources indicate that authorities are generally taking it slowly. Intelligence agencies are still at an early stage with AI agents, and the Defense Intelligence Agency has found that the underlying IT foundation must be in place before agents can be set to solve real tasks.

For ordinary readers, the point is that AI agents are making their way into large organizations, including the military. There, security and trust are often the decisive obstacle.