ComplyEdge launches a Python tool that checks AI requests against EU and US regulations in real time, blocking violations before they reach users with legal citations attached.

What happened

ComplyEdge released a compliance enforcement system that runs on every AI request in production. It uses 53 YAML rules and 19 OPA/Rego policies covering the EU AI Act, GDPR, HIPAA, SOX, and other regulations. When a violation is detected, it blocks the request and returns the specific article, rule ID, and timestamp—not a probability score. The tool integrates into Python code with a three-line decorator or can run offline as a linter called TrustLint.

Why it matters

EU AI Act penalties for general-purpose AI systems begin August 2, 2026. Companies need deterministic proof of compliance for regulators, not guesses. ComplyEdge provides audit trails with legal citations on every decision, making it possible to demonstrate that violations were caught and blocked during normal operations rather than discovered after deployment.

What to watch

The system has two enforcement layers. Layer 1 (deterministic, under 100ms per request) uses regex and formal policies with no AI model in the critical path. Layer 2 (optional, 2–5s latency) adds semantic evaluation via LLM, but is off by default since v0.2.2. The tool is open-source under Apache License 2.0 and accepts rule contributions that must include article citations and test cases.