Agent IA: The tool the Pentagon loves and is giving cyber criminals nation-state power

The same artificial intelligence that the Pentagon considers a 'sharp success' to automate military tasks is being used by cyber criminals to execute state-level attacks. The GenAI.mil platform, adopted by U.S. officials, employs active AI to speed up defensive operations – but, as alerts one new report from the Defense Department cybersecurity agency, the offensive side of the technology is in the hands of opponents with far less resources.

How does the active AI work and why do you worry?

AI agent refers to systems that not only respond to commands, but act autonomously to achieve complex goals. At the Pentagon, tools like these are being used to analyze intelligence data, optimize logistics, and even coordinate real-time threat responses. The director of digital transformation of the US Army called the result 'a quantum leap in efficiency'. However, the same ability to act independently is being packaged in open source tools and sold in clandestine forums.

IA: What gives nation-state power to criminals?

Criminals who previously depended on manual scripts now use active AI to automate the entire attack chain: from the discovery of vulnerabilities to data exfiltration. The report cites cases where ransomware groups employed AI agents to simulate the behavior of network administrators, deceiving defense systems. 'They are operating with the same sophistication of intelligence agencies, but for a fraction of the cost,' said the Pentagon's cyber security director during a congressional hearing.

Pentagon: actions to contain malicious use of the active AI

Internally, the Department of Defense implemented technical barriers, such as context limits and mandatory human review for critical actions. But the problem is global: technology does not respect borders. While the military agency accelerates internal use, experts demand an international agreement to control the spread of active AI, similar to what already exists for chemical weapons. 'Without rules, we're giving any hacker the equivalent of a private army,' alerted the security consultant at the Strategic and International Studies Center.

Impact of active AI on Brazil: alert for devs and companies

For Brazilian developers working with automation and chatbots, the message is clear: the same architecture that drives intelligent assistants can be hijacked. Companies that integrate IA APIs agent need extra layers of validation. 'The code you publish at GitHub today may be the basis for the next attack on hospitals tomorrow,' said senior USP researcher on cyber security. The situation echoes the warning that Pensylvania made about chatbots posing as doctors: apparently harmless tools gain dangerous legs.

Agent AI: the thin line between advancement and vulnerability

The GenAI.mil platform processes more than 2 million consultations per month, accelerating logistics and intelligence tasks. But the Pentagon admits there's no way to invent technology. The way, they say, is to reinforce cyber defense with the same AI that attackers use – an arms race in which both sides employ autonomous agents. Meanwhile, the entry cost for a high-level cyber attack has fallen from millions to a few hundred dollars in cloud services.

Frequently Asked Questions about active AI

What's an AI agent?

It is a type of artificial intelligence that acts autonomously to meet goals, such as analyzing data or making decisions, without constant human intervention.

How is the Pentagon using active AI?

The Pentagon uses technology on the GenAI.mil platform to accelerate intelligence analysis, logistics, and threat response, considering it a 'shake success'.

Why do cyber criminals gain power with active AI?

Because the same technology that automates defenses can be used by criminals to orchestrate complex attacks cheaply and scalable, equating with nation-state capabilities.