Artificial Intelligence for Services and National Security

Artificial Intelligence for Services and National Security
All Effects of Artificial Intelligence Use for Services and National Security

di Fabio Vanorio

What he said about Intelligence changes with the rise of Artificial Intelligence (AI) Anthony Vincey, Senior Fellow added to the National Security and Technology Program at the Center for a New American Security, already senior intelligence officer
In the long history of intelligence, the profession hasn’t changed much. Spies have always looked with their own eyes, listened with their own ears, analyzed and predicted with their own minds. A spy of Caesar’s legions and Lincoln’s Potomac Army would understand each other.
With the arrival of the machines, for the first time in the history of intelligence, a revolution is certain. And it is already underway, not everywhere in the world and, given the competitive specificity of the sector, there will be no catching-up process. Those who are not already equipping themselves today will succumb in the future.
At present, intelligence in countries of medium geopolitical stature is still based on the concept of a human operative physically following traces deemed relevant to national security,
and a human analyst connecting the dots to figure out what’s going on and try to predict what’s going to happen in the near future.
With the rise of Artificial Intelligence (AI) and autonomous systems, intelligence is radically transformed. The key points of the change emerged during a webinar held on May 5 at the Center for Security and Emerging Technology (CSET) of Georgetown University in which Anthony Vincey provided useful guidance.

• Vincey is a Senior Fellow added to the National Security and Technology Program at the Center for a New American Security, formerly a senior intelligence officer, serving with the Pentagon in Iraq, Africa and Asia in Operations, Ph. D in international relations at the London School of Economics, and member of the Council on Foreign Relations.
The essential elements of his intervention can be summarised in the following.

• The introduction of machines for the intelligence community consists of four elements. First, it changes what we spy on, therefore the objectives of intelligence. Second, it changes how we spy, therefore the mechanisms of collection. Thirdly, it changes the way we fight the others who spy on us, so counter-espionage. And finally, the mission of intelligence changes, so the surface of movement, traditionally composed of interactions between people.

• • Three technologies can be considered as directional vectors of the Revolution in Intelligence, ubiquitous sensors, AI and autonomy.

• • Autonomy is the most disruptive change since it removes humans from the intelligence cycle. Autonomy makes it possible to have differentiated entities that act in concert, allowing swarm operations. Swarms are a completely new form of combat, in which thousands or millions of vehicles operate simultaneously to overwhelm the enemy or fight other swarms. Human beings are not, and will not be, able to keep up with scale and complexity.

• • The conflict changes with the presence of autonomous systems/sensors in all domains (ubiquity) where human activities, from strategic planning to political decisions, develop and implement. As much as it may be desired that humans remain in the cycle, several factors limit what humans are able to understand in a world of autonomous drones teeming. Human beings are not able to keep up with the scale and complexity of an autonomous world of armament systems, of devices with self-development tactics that create models that no human can perceive with the speed of a machine, microseconds or nanoseconds, whether these are hypersonic missiles or attacks on financial markets.

• • In the world of autonomous warfare, espionage begins not at the time of the use of the devices, but already before their release, therefore in design, development, and production. Understanding the intent of designing such systems becomes an important part of the intelligence analysis process.

• • Autonomous systems are typically designed within a spectrum. What additional decisions the system can take alone compared to those involving a human operator becomes a factor of strategic knowledge about the opponent. Counter-intelligence activity therefore shifts to the development of autonomous systems. The analysis could reveal biases in the developer’s work that make it possible to inject codes into the system that weaken it before it goes out on the market. Software developers can, therefore, become new sources to recruit for the collection of information and secret actions against actors (state and not) hostile.

• • This necessary anticipation changes the process of recruiting sources that can no longer be carried out by agents with general knowledge of the computer environment (or worse, without a solid knowledge of the same). The recruiter will search the source for the specialization he needs at that time. This introduces the need for basic knowledge for the entire intelligence community, from the director to the scheduler, making abstruse the professionals who lack such skills.

• • Human intelligence (HUMINT) becomes even more important. What changes dramatically are the operations covered. Mass surveillance makes it virtually impossible to operate in hiding. The contrast of invasive surveillance, combined with fingerprints in social media, involves the use of other technical tools, such as models that confuse or manipulate computer vision systems.
• • HUMINT will require full integration with signal intelligence (SIGINT) and geospatial intelligence. In particular, the SIGINT, increased with AI, remains a milestone in the collection of information, as there will continue to be signatures of the electromagnetic spectrum and communications to be tracked and encrypted.
• • The adaptation of SIGINT also requires the removal of human beings from the intelligence cycle in order to manage, in a purely technical way, the entire scale and complexity of the operation. This is necessary to effectively collect data on autonomous systems and ubiquity sensors, also in view of the abnormal amount of geospatial “Big Data” impossible for humans to analyze on their own.
• The rule of compartmentalization becomes counterproductive. In a world dominated by a trust of autonomous systems, counter-espionage will require new protections, and significantly greater computer and economic expertise. Already today, the c.d. “poisoning of Machine Learning” is frequent, in which misleading data can be inserted inducing a machine learning system to recognize the wrong object (a computer vision system can be compromised in such a way as to exchange an enemy fighter jet for a passenger jet, and vice versa).
• • The technological transformation of the intelligence community does not take place confidentially, as in the Cold War. Instead, it goes in the opposite direction making intelligence less and less secret, with the presence of small amounts of highly classified information and most present in an unclassified space where will be, instead, the capabilities (and not the current bureaucratic security organizations) to constitute entry barriers. Competition becomes about knowledge, human and AI, about the advantage of innovation.

• • The technological transformation of the intelligence community does not take place confidentially, as in the Cold War. Instead, it goes in the opposite direction making intelligence less and less secret, with the presence of small amounts of highly classified information and most present in an unclassified space where will be, instead, the capabilities (and not the current bureaucratic security organizations) to constitute entry barriers. Competition becomes about knowledge, human and AI, about the advantage of innovation.

• • All individuals who take part in intelligence activity must be at ease with data science, and with technology in its various forms. It is an integrated man-machine system that continuously evolves to be more competitive. Competitive evolution becomes a natural state of national intelligence and security.

• The parallel between finance and intelligence is significant. In the protection of national security there are lives at stake, but for finance there are huge amounts of money where even small changes can mean a lot to the national economy. Both are highly technological, talent-based fields, and both are very averse to the risk of change.
• Given the parallels, focusing on finance helps predict the future of national security in an AI context. First of all, in finance, humans are already out of the loop. In high frequency trading, humans have adapted and accepted the exclusion from algorithmic “black boxes” at all levels and organizations, leaving (for now) only aspects of creativity to the human being. This aspect, especially in the way it has been implemented, can help to achieve a similar transition in the area of national security.
Fabio Vanorio is a director of the Ministry of Foreign Affairs and International Cooperation. He is an expert in Intelligence and National Security, as well as in military and security applications of Artificial Intelligence. He also writes for the Italian Institute of Strategic Studies “Niccolò Machiavelli”.
(All the opinions expressed are entirely by the author and do not reflect any official position attributable either to the Italian Government or to the Ministry of Foreign Affairs and International Cooperation)