Traccia: OpenTelemetry-Based AI Governance Platform Targets EU AI Act Compliance
New platform uses tamper-resistant telemetry to auto-generate compliance evidence for autonomous AI systems.
The rapid development of LLMs and AI agents has created a governance gap, as current monitoring tools (LLM evaluation platforms, ML workflows, APM systems) are disjointed and fail to protect against alignment drift, SaaS security threats, and unauthorized shadow AI deployments. International frameworks like the EU AI Act demand transparency and accountability, but practical implementations remain scarce. The paper identifies these shortcomings and proposes a unified solution.
Traccia builds on OpenTelemetry to provide a multi-level AI governance stack. It appends telemetry data, passive semantic guardrail assessments, and execution lineage into a hashed trace ledger with tamper-resistant fingerprints and SHA-256 content hashes. These compliance evidence packages automatically map to specific EU AI Act articles (12, 14, 19, 26(6), and 50) without compromising data privacy. The platform creates a machine-readable foundation for enterprise-wide management of autonomous AI systems, closing the gap between regulatory requirements and real-world deployment.
- Built on OpenTelemetry infrastructure for standardized telemetry collection across AI systems
- Auto-generates compliance evidence packages with SHA-256 hashes mapping to 5 EU AI Act articles (12, 14, 19, 26(6), 50)
- Addresses alignment drift, shadow AI, and SaaS security through execution lineage and passive semantic guardrails
Why It Matters
Provides a practical, standards-aligned framework for enterprises to govern autonomous AI systems transparently and compliantly.