AI in Military Operations: Challenges, Regulation, and Responsible Deployment

Artificial intelligence is primarily utilized by military intelligence services for surveillance, reconnaissance, and cognitive operations to shape public perception, according to Emelia Probasco, a Senior Fellow at Georgetown University’s Center for Security and Emerging Technology. Probasco distinguishes AI—which learns from data—from autonomous systems that rely on “if-then” logic.

How is AI currently used in military operations?

Military intelligence services use AI to capture the operative environment and interpret data from diverse sources. Probasco identifies cyberspace as a primary application area due to the massive volume of data generated there.

A significant impact is found in “cognitive operations.” According to Probasco, this involves analyzing media and social platforms to generate content that shapes how people perceive the world. Beyond this, AI improves sensor intelligence and network speed, providing long-term strategic advantages.

What is the difference between AI and autonomous systems?

The distinction lies in the underlying logic. Probasco explains that autonomous or automatic systems typically function on an “if-then” principle: when a specific event occurs, the system reacts in a predetermined way.

AI differs because it is driven by algorithms that learn from data, known as machine learning. Probasco cites Large Language Models (LLMs) like ChatGPT and Claude as primary examples of AI. These systems do not follow rigid “if-then” rules; they evolve based on the data they process.

Why isn’t AI fully integrated into drone warfare yet?

Despite frequent public discussion, Probasco states that technical problems currently prevent the broad introduction of AI in the drone sector. Most drones remain autonomous platforms steered by humans or basic autopilots rather than fully AI-driven entities.

How can AI be deployed responsibly in defense?

Responsible use requires a clear understanding of a system’s limits and capabilities. Probasco argues that operators must know where a system excels and where it fails.

She emphasizes the need for high-quality data and strict control processes. The goal, according to Probasco, should be to use AI to make better decisions rather than simply making decisions faster. Research from CSET suggests that the novelty of the technology makes this understanding a continuous challenge.

What challenges hinder the adoption of military AI?

Data quality is the primary obstacle. Probasco notes that AI cannot solve problems if the available data is in poor condition. Computational power also remains a bottleneck, as training and utilizing AI requires significant hardware resources.

Beyond technical hurdles, Probasco points to bureaucratic regulations and the need for better professional training. She suggests the technology is “good enough” for current purposes, but human proficiency in using it must catch up.

Should high-risk AI models be regulated?

Probasco advocates for the regulation of AI to ensure human safety. She specifically highlights the danger of models capable of discovering IT vulnerabilities, such as Anthropic’s Claude Mythos.

According to Probasco, models suspected of being harmful should not be released to the general public. Instead, she suggests these tools be vetted by governments or coalitions of companies and governments before any public availability to prevent societal harm.

What are the future trends for AI in the next five years?

Probasco positions herself between two extremes: those who believe compute laws have hit a ceiling and those who predict the arrival of Artificial General Intelligence (AGI) and “robot rulers.”

She expects the technology to continue making surprising leaps in capability. Probasco asserts that the most critical factor will be how humans decide which values to impose on these technologies. [Link to related article on AI Ethics]

Frequently Asked Questions