AI security experts discuss the growing risks of gray swan events where prompt injection and automated agent vulnerabilities threaten enterprise stability. They argue that traditional security models require specialized red-teaming tools to mitigate threats created by autonomous systems.