AI and National Security: Is the United States Doing Enough?

By: Roxanne Heston

Photo Credit: Getty Images

In recent years, the field of artificial intelligence (AI), a broad term used to discuss computer programs that can perform typically human functions, has emerged as a leading issue in the national security community. Newly piqued interest has prompted a flurry of activity on the topic, including a series of Congressional reports and a near doubling in Defense Department AI spending between 2015 and 2017.[i] While significant attention and investment have made the United States the preeminent nation for AI research and development thus far, current trends indicate that the U.S. may fall behind its competitors unless it improves its general attitudes toward AI integration and talent management.

AI already plays a significant role in defense and intelligence workplaces. For instance, at the National Geospatial Intelligence Agency (NGA), AI algorithms sift through vast troves of satellite footage; at the Office of Defense Strategy and Innovation, staff use AI to clean and structure assorted data types in support of counterterrorism efforts. While they are optimistic about the future of AI in their work, professionals also acknowledge the technology’s limitations. Some believe that machines exist to give “plausible possibilities,” not answers, and that algorithms suffer from the “failure to imagine” what attacks human adversaries might execute.[ii] This level of capability, however, is far from the vision that the Defense Advanced Research Projects Agency (DARPA) has for the future, one in which machines “function more as colleagues than as tools.”[iii] With that objective in mind, the United States’ strategy to maintain AI preeminence deserves scrutiny.

Many experts consider China to be the most significant non-Western nation in the AI research and development field, and likely the United States’ key competitor in the field.[iv] The Chinese national strategy published in 2017 sets a goal of becoming “the world’s ‘primary’ AI innovation center” by 2030, triggering substantial government funding to AI public-private partnerships and recruitment. However, the U.S. still clearly leads in private sector innovation, comprising a far greater fraction of the world’s hardware, developers, publications, and corporate investment.[v] By one estimate, China is doing only half as well as the U.S., lagging on all measures except volume of data, and even still limited by data sharing and integration.[vi]

Meanwhile, the U.S. Government has started new initiatives of its own. DARPA recently committed $2 billion to their AI Next Campaign to make systems more capable and robust.[vii] Existing activities like the Defense Innovation Unit (DIU) aid coordination with corporate developers, and the new Joint Artificial Intelligence Center will likely carry that vision further. The U.S. will also form an AI National Security Commission, striving to coordinate development efforts.[viii] Should it succeed, the U.S. stands to benefit from its unparalleled private and academic sectors, drawing not only from the domestic talent pool but also from the many international technologists flocking to those institutions.

While these activities certainly point towards continued American technological exceptionalism, some roadblocks still exist. The U.S. has yet to transform the NDAA push for greater industry cooperation into action, often due to the resistance of experts in the field. In a notable protest, last year Google employees withdrew from Project Maven, a Pentagon initiative to identify objects and people in surveillance footage.[ix] Signatories of a letter to the C.E.O. object to Google’s participation in “the business of war.”[x] Senior defense officials have said that they intend to avoid such incidents by being more transparent about what their projects are doing.[xi] Some express hope that the high stakes of national security work will act as a lure to overcome the salary disparity between the public and private sectors.[xii] Yet, this strategy is likely insufficient in a world where competitors offer hard cash instead of compelling narratives, and in which, as was the case with Google employees, the narratives themselves can fall flat.

The culture around AI development can also create certain challenges. The idea that AI is not meant to provide the “right answer” captures the general belief in security agencies that AI technologies are tools to augment existing processes, rather than potential agents to approach problems entirely differently.[xiii] Similarly, defense personnel often hesitate to fully automate processes, and tend to under-rely on automation for day-to-day activities.[xiv] This conservatism stems from the inconsistent performance of many current AI techniques and the vast risks preemptive deployment poses.[xv] But by continuing to hesitate on the development of reliable technologies, the U.S. will likely fall behind nations more willing to invest in long-term innovation.

While success depends on many inputs, talent is arguably the most valuable and difficult-to-acquire asset. At a minimum, the U.S. Government should offer faster and better paid government contracts for high-skill AI talent, snatching up those already interested in working on national security. For those who are not yet sympathetic, the government will need to create greater awareness of and affinity for defense and intelligence projects. For instance, as a House Subcommittee on Information Technology report suggests, a national competition like DARPA’s Grand Challenges would make it easier to “foster innovative, collaborative research.”[xvi] Lastly, while marketing is tricky and far from sufficient, its necessity is certain. The US government would benefit from communicating the defining nature of artificial intelligence in our time and the defining role U.S. national security investments will play in its development.












[i] U.S. Congress. House. Oversight and Government Reform. Rise of the Machines. By Will Hurd and Robin Kelly. 115th Cong., 2d sess. H. Rept. Washington, DC, 2018.

[ii] Brooks, Andrew, Valerie Browning, John Doyon, Meg Leta Jones, Patty Mims, and Eric Adams. “The New Age of Artificial Intelligence.” Proceedings of Kalaris Intelligence Conference, Georgetown University, Washington, DC.

[iii] United States of America. Defense Advanced Research Projects Agency. Department of Defense. DARPA.

[iv] “Global AI Market Revenue by Region 2016-2025 | Statistic.” Statista. 2018. Accessed October 06, 2018.

[v] Ding, Jeffrey. Deciphering China’s AI Dream. Report. Department of Philosophy, University of Oxford. March 2018.

[vi] Ibid.

[vii] United States of America. Defense Advanced Research Projects Agency. Department of Defense. DARPA.

[viii] Dura, Kathryn. “New Defense Policy a Reminder That US Is Not Alone in AI Efforts.” C4ISRNET. August 28, 2018. Accessed October 04, 2018.

[ix] Shane, Scott, and Daisuke Wakabayashi. “‘The Business of War’: Google Employees Protest Work for the Pentagon.” The New York Times. April 04, 2018. Accessed October 04, 2018.

[x] “Letter to Google C.E.O.” Letter to Sundar Pichai. In New York Times. April 4, 2018. Accessed October 4, 2018.

[xi] Browning and Jones, “The New Age of Artificial Intelligence”

[xii] Brooks, “The New Age of Artificial Intelligence”

[xiii] Scharre, Paul. Robotics On The Battlefield, Part I: Range, Persistence and Daring. Report. May 2014. ttp://

[xiv] Brooks et al., “The New Age of Artificial Intelligence”

[xv] Danzig, Richard. Technology Roulette: Managing Loss of Control as Many Militaries Pursue Technological Superiority. Report. June 2018.

[xvi] U.S. Congress, Rise of the Machines, 14

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.