Spy Vs. AI
U.S. Foreign Policy
Since its founding in 1922, Foreign Affairs has been the leading forum for serious conversation of American foreign policy and international affairs. The magazine has featured contributions from many prominent international affairs experts.
More Resources
- Feedback
- Institutional Subscriptions
- Gift a Subscription
- About Us
- Events
- Issue Archive
- Advertise
- Audio Content
- Account Management
- FAQs
Spy vs. AI
ANNE NEUBERGER is Deputy Assistant to the President and Deputy National Security Adviser for Cyber and Emerging Technology on the U.S. National Security Council. From 2009 to 2021, she served in senior operational functions in intelligence and cybersecurity at the National Security Agency, including as its first Chief Risk Officer.
- More by Anne Neuberger
Spy vs. AI
How Artificial Intelligence Will Remake Espionage
Anne Neuberger
-.
Copy Link Copied.
Article link: https://www.foreignaffairs.com/united-states/spy-vs-aihttps://www.foreignaffairs.com/united-states/spy-vs-ai.
Copy
Gift Link Copied.
This is a subscriber-only feature. Subscribe now or Check in.
Create Citation Copied.
Chicago MLA APSA APA.
Chicago Cite not available at the minute.
MLA Cite not available at the minute.
APSA Cite not available at the moment.
APA Cite not available at the moment
Download PDF.
This is a subscriber-only feature. Subscribe now or Sign in.
Request Reprint.
Request reprint approvals here.
In the early 1950s, the United States dealt with a critical intelligence challenge in its burgeoning competitors with the Soviet Union. Outdated German reconnaissance photos from World War II might no longer supply sufficient intelligence about Soviet military capabilities, and existing U.S. monitoring abilities were no longer able to penetrate the Soviet Union's closed airspace. This deficiency stimulated an audacious moonshot effort: the development of the U-2 reconnaissance aircraft. In just a couple of years, U-2 missions were delivering vital intelligence, capturing images of Soviet rocket setups in Cuba and bringing near-real-time insights from behind the Iron Curtain to the Oval Office.
Today, the United States stands at a comparable juncture. Competition between Washington and its competitors over the future of the international order is intensifying, and now, much as in the early 1950s, the United States need to make the most of its first-rate economic sector and adequate capacity for innovation to outcompete its adversaries. The U.S. intelligence community must harness the country's sources of strength to deliver insights to policymakers at the speed of today's world. The integration of expert system, particularly through large language models, provides groundbreaking opportunities to enhance intelligence operations and analysis, making it possible for the delivery of faster and more relevant assistance to decisionmakers. This technological revolution includes significant drawbacks, however, especially as foes exploit similar advancements to reveal and counter U.S. intelligence operations. With an AI race underway, the United States must challenge itself to be first-first to gain from AI, initially to secure itself from opponents who might use the technology for ill, and first to utilize AI in line with the laws and values of a democracy.
For the U.S. national security community, fulfilling the promise and managing the danger of AI will need deep technological and cultural changes and a willingness to change the way companies work. The U.S. intelligence and military neighborhoods can harness the capacity of AI while reducing its inherent dangers, guaranteeing that the United States maintains its competitive edge in a rapidly developing worldwide landscape. Even as it does so, the United States should transparently communicate to the American public, and to populations and partners around the world, how the country means to fairly and safely utilize AI, in compliance with its laws and worths.
MORE, BETTER, FASTER
AI's potential to reinvent the intelligence community lies in its ability to procedure and examine huge amounts of data at extraordinary speeds. It can be challenging to analyze big amounts of collected data to generate time-sensitive warnings. U.S. intelligence services might utilize AI systems' pattern acknowledgment abilities to determine and alert human analysts to prospective threats, such as rocket launches or military motions, or important international advancements that analysts know senior U.S. decisionmakers have an interest in. This capability would ensure that vital cautions are timely, bphomesteading.com actionable, and pertinent, allowing for more reliable actions to both quickly emerging threats and emerging policy opportunities. Multimodal designs, which integrate text, images, and audio, improve this analysis. For circumstances, using AI to cross-reference satellite images with signals intelligence could provide a detailed view of military motions, making it possible for quicker and more precise risk assessments and possibly new means of providing details to policymakers.
Intelligence analysts can also offload recurring and to machines to concentrate on the most satisfying work: creating initial and deeper analysis, increasing the intelligence neighborhood's general insights and efficiency. A fine example of this is foreign language translation. U.S. intelligence firms invested early in AI-powered capabilities, and the bet has paid off. The capabilities of language models have grown significantly sophisticated and accurate-OpenAI's recently launched o1 and o3 designs demonstrated considerable progress in precision and thinking ability-and can be utilized to much more rapidly translate and summarize text, audio, and video files.
Although difficulties remain, future systems trained on greater quantities of non-English data might be capable of discerning subtle distinctions in between dialects and understanding the meaning and cultural context of slang or Internet memes. By depending on these tools, the intelligence neighborhood might focus on training a cadre of highly specialized linguists, who can be difficult to discover, typically battle to survive the clearance process, and take a long time to train. And obviously, by making more foreign language materials available across the ideal agencies, U.S. intelligence services would be able to quicker triage the mountain of foreign intelligence they get to choose the needles in the haystack that really matter.
The value of such speed to policymakers can not be underestimated. Models can swiftly sort through intelligence information sets, open-source details, and traditional human intelligence and produce draft summaries or preliminary analytical reports that experts can then validate and improve, making sure the final products are both detailed and precise. Analysts could team up with a sophisticated AI assistant to resolve analytical issues, test concepts, and brainstorm in a collaborative fashion, improving each model of their analyses and delivering ended up intelligence faster.
Consider Israel's experience in January 2018, when its intelligence service, the Mossad, discreetly got into a secret Iranian facility and took about 20 percent of the archives that detailed Iran's nuclear activities in between 1999 and 2003. According to Israeli officials, the Mossad gathered some 55,000 pages of documents and a further 55,000 files kept on CDs, consisting of pictures and videos-nearly all in Farsi. Once the archive was obtained, senior officials put tremendous pressure on intelligence experts to produce detailed assessments of its material and whether it indicated a continuous effort to build an Iranian bomb. But it took these professionals a number of months-and hundreds of hours of labor-to translate each page, examine it by hand for appropriate material, and include that details into assessments. With today's AI abilities, the first 2 steps in that process could have been achieved within days, possibly even hours, enabling analysts to understand and contextualize the intelligence rapidly.
Among the most interesting applications is the method AI could change how intelligence is taken in by policymakers, allowing them to engage straight with intelligence reports through ChatGPT-like platforms. Such capabilities would enable users to ask particular questions and get summed up, relevant details from countless reports with source citations, assisting them make informed choices rapidly.
BRAVE NEW WORLD
Although AI uses many advantages, it likewise postures significant brand-new risks, particularly as foes develop comparable technologies. China's improvements in AI, especially in computer vision and surveillance, threaten U.S. intelligence operations. Because the nation is ruled by an authoritarian program, it lacks privacy constraints and civil liberty defenses. That deficit makes it possible for massive data collection practices that have actually yielded data sets of enormous size. Government-sanctioned AI designs are trained on vast amounts of personal and behavioral data that can then be utilized for various purposes, such as monitoring and social control. The existence of Chinese business, such as Huawei, in telecommunications systems and software application around the globe could offer China with prepared access to bulk information, notably bulk images that can be utilized to train facial recognition designs, a specific issue in countries with large U.S. military bases. The U.S. nationwide security community need to consider how Chinese designs developed on such comprehensive information sets can give China a strategic advantage.
And it is not just China. The proliferation of "open source" AI designs, such as Meta's Llama and those created by the French company Mistral AI and the Chinese company DeepSeek, is putting powerful AI abilities into the hands of users around the world at fairly budget-friendly expenses. A lot of these users are benign, however some are not-including authoritarian programs, cyber-hackers, and criminal gangs. These malign stars are utilizing big language models to rapidly create and spread incorrect and destructive content or to conduct cyberattacks. As witnessed with other intelligence-related technologies, such as signals obstruct capabilities and unmanned drones, China, Iran, and Russia will have every reward to share some of their AI developments with client states and subnational groups, such as Hezbollah, Hamas, and the Wagner paramilitary company, therefore increasing the threat to the United States and its allies.
The U.S. military and intelligence community's AI designs will become appealing targets for adversaries. As they grow more powerful and main to U.S. national security decision-making, intelligence AIs will become vital national possessions that need to be protected against adversaries looking for to compromise or manipulate them. The intelligence neighborhood must buy establishing protected AI designs and in establishing requirements for "red teaming" and constant evaluation to secure against potential hazards. These groups can utilize AI to simulate attacks, revealing possible weaknesses and developing techniques to mitigate them. Proactive procedures, consisting of collaboration with allies on and financial investment in counter-AI innovations, will be necessary.
THE NEW NORMAL
These difficulties can not be wished away. Waiting too long for AI innovations to completely mature brings its own risks; U.S. intelligence capabilities will fall behind those of China, Russia, and other powers that are going full steam ahead in establishing AI. To make sure that intelligence-whether time-sensitive cautions or longer-term strategic insight-continues to be a benefit for the United States and its allies, the country's intelligence community needs to adjust and innovate. The intelligence services need to rapidly master the usage of AI technologies and make AI a fundamental element in their work. This is the only sure way to guarantee that future U.S. presidents receive the finest possible intelligence support, remain ahead of their adversaries, and protect the United States' sensitive abilities and operations. Implementing these modifications will require a cultural shift within the intelligence neighborhood. Today, intelligence analysts mainly construct items from raw intelligence and data, with some assistance from existing AI designs for voice and imagery analysis. Moving forward, intelligence authorities should explore consisting of a hybrid technique, in line with existing laws, utilizing AI designs trained on unclassified commercially available information and fine-tuned with classified details. This amalgam of technology and traditional intelligence gathering could result in an AI entity offering direction to images, signals, open source, and measurement systems on the basis of an incorporated view of regular and anomalous activity, automated images analysis, and automatic voice translation.
To accelerate the transition, intelligence leaders need to promote the advantages of AI combination, stressing the enhanced abilities and performance it provides. The cadre of newly designated chief AI officers has been established in U.S. intelligence and defense to function as leads within their firms for promoting AI development and getting rid of barriers to the innovation's execution. Pilot tasks and early wins can develop momentum and confidence in AI's capabilities, motivating more comprehensive adoption. These officers can utilize the know-how of nationwide laboratories and other partners to evaluate and refine AI designs, guaranteeing their effectiveness and security. To institutionalise change, leaders should develop other organizational incentives, consisting of promos and training opportunities, to reward inventive techniques and those workers and units that demonstrate reliable use of AI.
The White House has actually created the policy required for using AI in national security companies. President Joe Biden's 2023 executive order relating to safe, protected, and trustworthy AI detailed the guidance required to fairly and securely use the technology, and National Security Memorandum 25, released in October 2024, is the nation's fundamental technique for harnessing the power and managing the dangers of AI to advance nationwide security. Now, Congress will require to do its part. Appropriations are needed for departments and agencies to create the facilities needed for development and experimentation, conduct and scale pilot activities and assessments, and continue to invest in examination capabilities to ensure that the United States is constructing dependable and high-performing AI technologies.
Intelligence and military neighborhoods are committed to keeping people at the heart of AI-assisted decision-making and have actually created the frameworks and tools to do so. Agencies will need standards for how their analysts must use AI models to make certain that intelligence items fulfill the intelligence neighborhood's requirements for dependability. The government will also require to maintain clear assistance for handling the information of U.S. people when it pertains to the training and use of big language models. It will be necessary to stabilize the usage of emerging technologies with securing the privacy and civil liberties of residents. This indicates augmenting oversight mechanisms, updating relevant structures to show the capabilities and risks of AI, and fostering a culture of AI advancement within the nationwide security device that utilizes the capacity of the technology while safeguarding the rights and freedoms that are foundational to American society.
Unlike the 1950s, when U.S. intelligence raced to the forefront of overhead and satellite images by developing a lot of the key technologies itself, winning the AI race will need that community to reimagine how it partners with private market. The economic sector, which is the main means through which the government can understand AI progress at scale, is investing billions of dollars in AI-related research, data centers, valetinowiki.racing and calculating power. Given those companies' improvements, intelligence companies should focus on leveraging commercially available AI designs and refining them with classified information. This approach makes it possible for the intelligence neighborhood to rapidly expand its capabilities without needing to go back to square one, permitting it to remain competitive with enemies. A recent partnership between NASA and IBM to develop the world's biggest geospatial foundation model-and the subsequent release of the design to the AI community as an open-source project-is an exemplary presentation of how this type of public-private collaboration can operate in practice.
As the nationwide security neighborhood integrates AI into its work, it must make sure the security and strength of its designs. Establishing requirements to deploy generative AI firmly is important for maintaining the stability of AI-driven intelligence operations. This is a core focus of the National Security Agency's new AI Security Center and its partnership with the Department of Commerce's AI Safety Institute.
As the United States faces growing rivalry to shape the future of the worldwide order, it is urgent that its intelligence agencies and military capitalize on the country's innovation and management in AI, focusing especially on big language models, to offer faster and more pertinent details to policymakers. Only then will they gain the speed, breadth, and depth of insight required to browse a more intricate, competitive, and content-rich world.