Combatting Generative AI Threats

As we have seen, incorporating AI into cybersecurity provides great benefits but comes with significant risks and vulnerabilities. Fortunately, there are security controls that can help ensure that the AI-enabled systems we deploy for cybersecurity are safe and secure. This chapter will shift the focus away from developing and deploying safe and secure AI systems to some other risks and threats related to AI use. With the emergence of generative AI and LLMs, AI use is becoming commonplace. The incredible power of AI is now in the hands of the general public, and organizations are racing to find how to harness this power within a multitude of current and novel processes. However, bad actors are also anxious to unleash the power of generative AI and LLMs to commit fraud and other cybercrimes. Just like with the development and deployment of AI systems, we must understand the threats. This chapter examines some of the ways LLMs are used for malicious purposes, including dark LLMs, AI jailbreaking, deepfakes, and disinformation. Other concerns related to LLMs, such as bias and privacy, were addressed in Chapter 1 .