Insider Insights: The Intersection of Artificial Intelligence and Cybersecurity in Military Technology

In a recent interview with The Associated Press, Shift5 founder Josh Lospinoso discussed the role of artificial intelligence (AI) in cybersecurity and military technology. The former Army captain and Rhodes Scholar outlined how software vulnerabilities in weapons systems pose a major threat to the security of the U.S. military. As technology advances, software-enabled systems have become more interconnected and porous, making them vulnerable to attacks by malicious actors. AI can help protect these systems, but it can also be used by hackers to try and defeat detection and classification systems.

Software vulnerabilities in weapons systems

Lospinoso’s first concern is the existing software vulnerabilities in weapons systems. According to a 2018 Government Accountability Office report, nearly all newly developed weapons systems had mission critical vulnerabilities. When attackers gain access to these systems, it is game over. Lospinoso argues that fixing these vulnerabilities is a technical debt that will take a long time to pay. Because many of these systems are decades old and have been retrofitted with digital technology, it may be easier in some cases to build new platforms than redesign existing systems.

The new frontier: securing AI algorithms

The second concern is securing AI algorithms in these systems. As AI development accelerates, Lospinoso argues that there is a risk of rushing AI products to market without building in adequate security measures. This “burning-use” case, where a product is rushed to market and catches fire (gets hacked, fails or causes unintended damage), could lead to significant economic dislocations and a fundamental change in business models. Lospinoso emphasized that building secure and responsible AI products is crucial to avoid unintended damage or harm.

Data poisoning: a major threat to AI-enabled technologies

Lospinoso also discussed the two principal threats to AI-enabled technologies. The first is theft, which is self-explanatory, and the second is data poisoning. Lospinoso describes data poisoning as digital disinformation. If adversaries are able to craft the data that AI-enabled technologies see, they can profoundly impact how that technology operates. An example of this occurred in 2016 when Microsoft released a Twitter chatbot called Tay that learned from conversations it had online. Malicious users conspired to tweet abusive and offensive language at it, causing Tay to generate inflammatory content.

The role of AI in cybersecurity and military technology

AI has been integral to the field of cybersecurity for some time. AI is used in email filters to flag junk mail and phishing lures, as well as in malware detection software that runs on networks. However, hackers also use AI to try and defeat these classification systems, a phenomenon known as adversarial AI. Lospinoso argues that AI can be used to defend against attacks on military software systems and can help improve maintenance and operational intelligence. Nonetheless, he believes that AI algorithms are not yet ready for prime time when it comes to military decision-making, such as targeting.

Advice for advancing AI research

Finally, Lospinoso addressed concerns that some have urged pausing AI research. He believes that such a pause would be a mistake, as it would favor China and other competitors. However, he does caution that rushing AI products to market without adequate security measures could lead to unintended consequences and long-term damage. He recommends that AI research should move forward with a focus on building secure and responsible products, rather than racing to market with half-baked products. Lospinoso emphasized the need for continued discussion and collaboration between government and business leaders to address issues of AI development and cybersecurity in military technology.