Much of the discussion around emerging generative artificial intelligence (GenAI) tools has focused on their potential to be used to undermine stability, security and trust in institutions through disinformation, automated uncrewed weapons systems and deepfakes. 


However, a lesser-discussed dimension of AI is these same tools' potential to serve valuable defense purposes, including threat detection, enhanced decision-making and large-scale data analysis.


OpenAI's Defence Contract
OpenAI, a leading U.S.-based AI firm, recently reported working with the U.S. Department of Defense (DoD) to strengthen the latter's AI capabilities. 


This highlights militaries' eagerness to adopt AI and dependence on the private sector to augment internal innovation or fill gaps left by the shortcomings of internal efforts.


Many militaries, including the US Armed Forces, have worked on developing AI capabilities internally over the past decade:

• In 2020, the DoD adopted a set of “Ethical Principles for Artificial Intelligence”; in 2022, it established the Chief Digital and Artificial Intelligence Office.

• Before that, in 2017, the U.S. military established Project Maven, its most publicly visible AI initiative intended to help integrate machine learning tools into military systems for scanning surveillance footage and other applications.


Initially supported by a contract with Google, that project proved controversial with the company's employees. It was later transferred to the National Geospatial Agency.


Google is not the only technology company to face internal tensions about working with military customers. OpenAI itself previously banned applications of its services for “military and warfare” purposes. 


The company has since revised the policy to exclude only services for “weapons, destroying property and harming people.”


Military and other government contracts are invaluable for AI companies looking to grow rapidly and fund the massive computing power necessary for the models underlying their services.


Prime Military Uses
Not surprisingly, militaries tend to be secretive about how they use or intend to use AI tools. However, several application areas seem ripe for incorporating AI, including:

• scanning surveillance imagery and videos of suspicious activities or other notable events, and for the detection of anomalies;

• automating the delivery of logistics and supply chains for military operations; and

• supporting decision-making by integrating analysis of multiple different data streams.


Such functions would meet OpenAI's criteria for not directly facilitating weapons development or harming people and property while enabling the company to support large contracts with military customers.


Geopolitical Competition
Such contracts will likely intensify competition between international rivals, especially the United States and China, to cultivate domestic AI companies and avoid relying on foreign firms for AI-powered defense technologies.


While many countries (including China) are focused on developing a robust AI industry for economic applications, expanding the range of military applications will provide further impetus to national investments and influence the direction of private-sector innovation.


The use of AI to develop lethal autonomous weapons systems (LAWS) is another critical issue for the defense sector. There are heightened concerns about the ability of machines to track, identify and execute targets without human intervention.


These are the types of weapons that OpenAI and many other private sector companies have largely resolved not to work on. 


That does not preclude the possibility of militaries developing these themselves. There are early reports of deployments of such systems in conflicts, including in Gaza, Ukraine and Libya.


From a technical perspective, the most challenging component of such systems is the identification and verification of target—the actual firing of weapons generally does not require sophisticated machine learning.


This means that private firms helping military stakeholders improve their image analysis and computer vision capabilities could, indirectly, feed into the development of better, more accurate LAWS.


U.S.-Chinese Competition
The United States is reckoned to lead China in AI capabilities, if only by a small margin. However, this advantage comes primarily from private-sector innovation and does not necessarily extend to military or government advancements.


A study sponsored by the Center for Security and Emerging Technology found that from April to November 2020, the Chinese and the US militaries invested heavily in AI procurement contracts, particularly systems for autonomous vehicles, intelligence-gathering, surveillance and reconnaissance. 


The report found that the two militaries were devoting comparable resources to AI contracts and focusing their investments in similar areas.


It also found that the militaries relied on a relatively distributed network of vendors, many of whom were smaller than traditional military contracting firms. 


This suggests that militaries are relying more on start-ups and newer technology firms for these contracts rather than just using the large contractors with whom they have built relationships over many years.


Willingness to contract with new partners may also reflect militaries' desire to buy the most recent emerging AI technologies. These come from newer companies rather than legacy defense firms that are not necessarily at the cutting edge of AI innovations.


That pressages a shift in the investment landscape of defense contracting.