AI Development in Dire Need of Cybersecurity Regulations

Artificial intelligence drives forward not only the biggest industries, but also the biggest concerns for cybersecurity in general. Science fiction authors and specialists have long speculated about the possible impact of the future technology. Now that it is finally available across the world we can state that the fictions realms are no longer just a product of the imagination. 

AI Has Become One of the Most Important Technologies 
The global superpowers are already on a race, one noteworthy example is China’s plans to dominate artificial intelligence development. The upcoming technology will not only replace real humans and their jobs by automating tasks, but also take decisions based on algorithms that can actively make a difference on a global scale.

The real concern is how the AI will be able to defend itself from threats. While some of the basic types of intelligent agents are already used in factories across the world, the AI of the future will be immeasurably superior. Even though the big corporations that implement such technology in their facilities spend large sums of money for security purposes, we witness critical hacker intrusions almost every week. World leaders are now putting cybersecurity defense as one of the most important topics of their agenda. Having this in mind, the development and adoption of secure standards for AI security is of utmost importance.  

By considering the facts that a hacker intrusion or enemy sabotage of an artificial intelligence system can lead to modifications of the AI “core”, we must take into account that human lives can be at stake. As the technology matures it will become standard to replace human controllers and operators with automated workers for the simpler tasks. It would not take long before AI actions will have a large impact globally. This is the reason why a thorough and in-depth revision of all current and future cybersecurity issues must be taken in account well before any AI systems are deployed in production use.

AI Cybersecurity Must Be both Preventative and Proactive
A rampant or hacked artificial intelligence can take over systems and networks and even interact with other similar agents to perform malicious actions. Depending on their ability to interact with the surrounding world, they can cause problems on a global scale. By design the intelligent agents are autonomous and if they "break away" from the prescribed "rules", it can be hard to predict the exact outcome of their actions. 

Security infiltrations, hacker attacks and vulnerability testing against artificial intelligence beings can happen and if the criminals start their assault using large resources, then a successful attack can become reality. 

Now is the time to act! Strict security standards have to be drafted and implemented even in the simplest forms of AI. This is the starting point from where future improvements to the documents will have to be made. AI like all other "digital man-made" technologies has the potential to inflict a lot of damage if security breaches to any of its modules are caused. The implications of such malicious actions can have dire consequences upon the world, and depending on the case they may be beyond the control of the human operators.

To prevent a global catastrophe, experts from different fields must focus on hacking prevention. A proactive approach relies on the ability to respond effectively against hacker attacks. The AI systems must be equipped with the means of knowing for themselves what a malicious action is.  To implement such a “feature”, the engineers must create detailed specifications to govern how the agents will behave when an intrusion is attempted. Such publications can be created similar to the Request for Comments (RFC), the memorandum documents that regulate the Internet. 

Discussions on this topic should continue further with specific suggestions and even draft proposals. A reasonable way of going forward in this direction is by identifying common concerns among government institutions, academia and the leading industry vendors. Every year hundreds of specialist security conferences, reports and meetings are planned. It would not be difficult to take the most popular issues that can make up the initial agenda of a proposed draft submission. 

The IEEE has already started working on a global initiative regarding AI planning by creating several documents on the topic. It is important to note that the first versions emphasize on "human well-being" and safety prescriptions when working with artificial superintelligence.

Hopefully such actions would commence soon, as the AI development industry is rapidly expanding. In our real-world case any delays can have fatal consequences.

Martin Beltov graduated with a degree in Publishing from Sofia University. As a cybersecurity enthusiast he enjoys writing about the latest threats and mechanisms of intrusion.

What’s Hot on Infosecurity Magazine?