Sign up to receive free updates on Technology.
We will send you a myFT Daily Digest email every morning, summarizing the latest news in the field of Technology.
The writer is the former director of cyber at GCHQ and now works as a consultant on technology and leadership.
In a recent speech, Sir Richard Moore, the head of UK intelligence agency MI6, highlighted the importance of the “human factor” in espionage, despite the potential benefits of artificial intelligence (AI). However, the rapid advancements in AI could greatly enhance the operations of intelligence agencies in response to growing security threats. The question arises: how can AI be integrated into national security while preserving the crucial human element?
At its core, AI excels in pattern recognition, capable of learning and generating patterns at incredible speed with reliable training data. It can assist analysts in processing information, identifying trends, and aiding in decision-making. GCHQ, a leading intelligence agency in the UK, has already harnessed AI to support human investigators by identifying and intercepting imagery, messages, and connections to locate offenders. Additionally, AI holds promise in uncovering hidden people and illegal activities on the dark web.
Intelligence agencies, whether focused on human recruitment or communication interception, are dedicated to making sense of complexity. They interpret events using fragmented data to provide insights that guide foreign policy and crisis response. In such a task that often involves finding the proverbial needle in a haystack of information, AI undoubtedly has a role to play.
Nevertheless, there are significant challenges ahead. Western intelligence agencies, like the Five Eyes alliance countries (US, UK, Canada, Australia, and New Zealand), highly value their reputation as trusted entities operating within legal, ethical, proportionate, and responsible frameworks. Balancing this reputation while adapting to new digital capabilities remains a major challenge.
AI faces difficulties in complex and evolving scenarios. Its safe and secure development, even in controlled environments like hospitals, has been a formidable task. Searching for unseen elements in contested environments where adversaries intentionally obscure information is still a work in progress.
AI excels in analyzing the “what most people would do” types of scenarios but struggles with outliers. Yet, identifying outliers is the core focus of the intelligence community, particularly when searching for extremists rather than average citizens. AI is exceptional at extrapolating from known data but struggles with determining the unknown or unseen.
Moreover, AI currently lacks contextual understanding, the ability to correct erroneous input, and the capacity to verify the reliability of its source information. To an AI system, a picture of a burning building with today’s date is treated as an ongoing event, not a historical artifact. Statements about climate change from online debates are considered equally valid, and AI cannot differentiate between accurate information and wild inaccuracies without being specifically programmed to do so, which may introduce biases.
Not only are AI data inputs susceptible to human interference, but its responses are also vulnerable to manipulation. Just as the codebreakers and codemakers at GCHQ’s Bletchley Park during World War II faced challenges, AI is also subject to such disruptions. Even the most advanced driverless vehicles in San Francisco can be stalled by pranksters placing traffic cones on them. AI lacks the instinctive response of a human when encountering unforeseen events.
In the UK, researchers in science and industry are actively investigating AI safety and security to establish robust standards for the future. They seek to incorporate features that counter adversarial practices into AI design from the initial stages. The threat of poisoning an AI system through corrupt training data remains a critical concern.
Undoubtedly, AI’s most obvious application, for intelligence agencies and numerous other professions, lies in managing the vast amounts of data produced in today’s world. Our contemporary society generates more data per minute than the entire Ancient Greek civilization stored. However, the challenge lies in distinguishing credible information from the sea of misinformation constantly disseminated by adversaries. This is why intelligence agencies, like any profession pursuing the truth, still require human involvement.
Sir Richard Moore stated that as AI delves into the vast ocean of open-source data, the unique characteristics of human agents in strategic positions will become increasingly significant. Human agents are not passive collectors of information; they can actively influence decisions within governments or terrorist organizations. However, there is currently no automated shortcut to replicate this human influence.
Denial of responsibility! VigourTimes is an automatic aggregator of Global media. In each content, the hyperlink to the primary source is specified. All trademarks belong to their rightful owners, and all materials to their authors. For any complaint, please reach us at – [email protected]. We will take necessary action within 24 hours.
