Generative AI Attacks Require Dynamic Defense

Generative AI technology are afraid of losing control over Generative AI systems themselves. That is, Generative AI systems acting like the sorcerer’s apprentice. This can be as simple as receiving unexpected results, or as dangerous as actions that threaten lives. Some go so far as to be concerned about the survival of the human race. Governments are taking these concerns very seriously.

Dynamic Defensive Tools

To defend against Generative AI created attacks, we have to develop alternatives to the sieve and the recipe. We need ways of recognizing that an attack is underway that do not depend on patterns. Similarly, we need ways of determining remediation without relying on recipes. One way is to follow the dynamic and adaptive methods nature has used in biology, but adapted to digital information systems. Because of:

• Latency issues, they must work at the point of attack;
• Remediation involving more than one part of a system, they must enable cooperation between system components for recovery;
• Volume of information problem, they must have only the data needed, where it is needed, when it is needed;
• Concerns about systems getting out of human control, they must have flexible ways for manual intervention to guard against unexpected outcomes.
  

These dynamic and adaptive defenses would be able to dynamically identify attacks and dynamically determine and implement remediation. They can be characterized as D2 systems.

S2-D2 Defenses

It is tempting to imagine a scenario where we move rapidly from our current S2 defenses to D2 defenses. D2 systems can defend against S2 attacks as well as D2 attacks. But, organizations have large sunk investments in S2 defenses. Attackers also have large investments in S2 attack systems. We are beginning to see an upsurge in Generative AI-created attacks. But, there continue to be large volumes of static attacks. Thus, it makes sense to have defenses that can take advantage of the sunk investments in S2 defenses while also fielding D2 defenses. Such systems can be characterized as S2-D2.  It is these S2-D2 systems we need to quickly develop.

Responding to the Challenge

Western governments need to band together with innovators in commercial industries to support R&D focused on the rapid development of S2-D2 systems. Governments in the US, Europe, Canada, Australia, and New Zealand have programs to provide financial support for R&D and entrepreneurship. These programs come from both the national defense and the industrial policy departments of these governments. Because of this, they have tended to be narrow in focus—often more concerned with improving national competitive positioning against the rest of the world. But, the threat from Generative AI cyberattacks is a global one. The Western world needs an effective response—and quickly. This can best be achieved through established cooperation between the leading Western governments while working closely with innovators in the commercial industry.