3 min read
Vulnerabilities A critical deserialization vulnerability (CVSS 9.8) in a widely-deployed ML model serving framework allows unauthenticated RCE via crafted model files. Active exploitation confirmed.
All Articles
5 min read
Research Safety-aligned multimodal LLMs can be reliably jailbroken by encoding adversarial instructions as text within images, bypassing text-layer safety filters entirely.
5 min read
Defensive Techniques Applying zero trust to ML infrastructure: training pipeline access controls, model registry security, inference endpoint hardening, and secrets management for AI deployments.
5 min read
Incident Reports An NHS trust confirmed adversarial perturbations applied to medical images caused systematic misclassification by its AI diagnostic system, resulting in incorrect preliminary diagnoses.
5 min read
Incident Reports A UK insurer's AI chatbot, due to an IDOR vulnerability and excessive tool permissions, allowed authenticated users to retrieve policy data for unrelated customers. 80,000 records exposed.