Could your own AI agents run a ransomware attack?

Ransomware is evolving well beyond locking systems, and agentic AI is introducing a category of security risk most organizations are not yet equipped to handle.

On The Cybersecurity Defenders Podcast, Behnaz Karimi, Senior Cybersecurity Analyst at Accenture and independent ransomware researcher, walks through what that shift actually looks like.

The full conversation includes:

• Why ransomware is increasingly targeting AI models, training data, and inference pipelines rather than just locking systems
• How data poisoning attacks make it difficult to determine when a model was compromised and whether its outputs can be trusted
• What multi-agent architectures introduce as an attack surface through inter-agent communication
• Why prompt injection has no complete solution and what that means for agentic deployments Why threat modeling for AI systems needs to happen before deployment, not after