Enterprise AI Governance

Start with discovery: shadow AI (unsanctioned tools and models) is the ungoverned surface — you can't secure what you don't know exists.

How to detect, govern, and secure unauthorized AI tool use in the enterprise — before sensitive corporate data leaves through AI prompts.

The regulatory landscape: EU AI Act, NIST RMF, and ISO 42001 mapped to practical controls — what you must do and by when.

The EU AI Act creates statutory security obligations for high-risk AI systems; NIST AI RMF 1.0 and ISO/IEC 42001:2023 are becoming de facto requirements through procurement and certification. This post maps specific requirements — adversarial testing, incident reporting deadlines, robustness mandates — to the attack classes they govern and the controls that address them.

Different models carry different risk profiles — this post frames how to assess and document the security posture of each model in your stack.

We ran the same security probe set against three models behind the same Copilot CLI wrapper. The wrapper never changed. The refusal behavior did — significantly.

Production monitoring: what to log, what anomalies to alert on, and how to build an audit trail for LLM API interactions.

Guardrails block known bad inputs. Incident response handles breaches after they happen. This post covers the gap: building LLM-specific observability that detects attacks while they are occurring — anomalous outputs, prompt injection signatures, unusual tool-call patterns, and model drift from baseline.

A decision framework for guardrail design — when to use input filters, output classifiers, or structural controls, and how to avoid the fragile middle.

LlamaGuard, Azure Prompt Shield, PromptGuard, NeMo Guardrails, Guardrails AI, and Presidio — how each works and when to deploy it.

Why guardrails fail: the structural bottleneck problem — a single guardrail layer is a single point of failure regardless of its accuracy.

New research finds that guardrail performance on tool-call trajectories correlates at ρ=0.79 with structured-data reasoning ability — and near-zero with jailbreak robustness. Here's what that means for how you secure agents.

When researchers find vulnerabilities in your models: how CVD processes for AI differ from traditional software and what an effective programme looks like.

Traditional CVD is well-established for software vulnerabilities. AI models break nearly every assumption it rests on. Here's where the security community stands — and what a workable disclosure framework might look like.

Caps the path: the incident response playbook — detection, containment, eradication, and post-incident review specific to AI security events.

Detection signals, containment options, evidence preservation, and recovery procedures for AI-specific security incidents — the operational complement to attack coverage. NIST IR lifecycle applied to prompt injection, model backdoors, data poisoning, and adversarial input attacks.