# What should security leaders do with AI? They don’t know.

## Executive summary

Cybersecurity leaders face 'AI decision fatigue' due to the rapid pace of AI adoption and emerging threats. To mitigate this paralysis, experts recommend starting AI deployment with low-risk, repetitive tasks (L1/L2 alerts) or by arming red teams. The discussion highlighted sophisticated new attack vectors like Ghostjacking—which poisons trusted data sources (logs, alerts)—and questioned the reliability of AI in patching vulnerabilities. Overall, successful AI integration requires a 'human-in-the-loop' approach and strict adherence to Zero Trust principles, particularly regarding agent permissions.

## Key takeaways

- Start AI deployment with repetitive tasks: Initial AI adoption should focus on automating predictable, high-volume activities (e.g., triaging L1/L2 alerts or vendor risk assessments) to reduce staff fatigue and build confidence in the technology.
- Red Teaming is key for defensive AI use: Equipping red teams with advanced tools allows them to simulate sophisticated threat actor behavior, enabling defenders to proactively develop countermeasures against emerging attack methods.
- Ghostjacking targets trusted data sources: A new attack method involves sneaking malicious prompts into highly trusted systems like logs, alerts, and error reports. This is a sophisticated evolution of prompt injection that bypasses initial security controls.
- AI patching requires human oversight: Research indicates AI-generated patches (e.g., using GPT 5.5) have a low success rate, often failing to solve the underlying vulnerability or creating new issues. Human validation and partnership are crucial.

## Technical details

- Ghostjacking Attack Vector: This technique is a sophisticated prompt injection that embeds malicious commands into highly trusted data sources (logs, alerts, error reports). An example cited was an agent reading a log recording a blocked connection request and being compromised by the embedded prompt.
- AI Patching Efficacy: Research showed that when researchers generated 540 patches for six vulnerabilities using GPT 5.5 and Opus 4.8, only 46% solved the underlying vulnerability, suggesting AI is not a perfect replacement for human due diligence.
- Agentic Identity Management: To mitigate risks from autonomous agents, it is critical to implement strict Zero Trust principles by limiting agent permissions and requiring a 'human-in-the-loop' for high-risk maneuvers (e.g., privilege elevation or command submission).

## Practical implications

- Adopt a phased approach to AI implementation, starting with automating repetitive tasks (L1/L2 alerts) where the process is well-understood.
- Treat all automated agents as potentially compromised and enforce strict least-privilege access controls. Never remove human oversight entirely.
- When developing or testing code patches using AI, assume failure and implement rigorous multi-agent validation processes to ensure security and stability.
- Focus on defining the desired business outcome first, rather than selecting an AI tool, to avoid decision paralysis.

## Topics

AI Security, Prompt Injection, Cybersecurity Strategy, Agentic Systems, Patch Management, Zero Trust Architecture, IBM X Force Cyber Range, Tenet Security DEFCON Presentation

Source: https://www.youtube.com/watch?v=3zhveTN1VzE