Distributed Tracing Shows You What Happened. It Cannot Prove It to a Regulator.

Short summary

OpenTelemetry traces provide engineering visibility into AI agent execution but fail as regulatory evidence under EU AI Act Articles 9 and 13. The core problem is structural: self-reported telemetry is the system auditing itself, which regulators will not accept as proof. Three failure modes—mutable telemetry, vendor-controlled storage, and context window gaps—cannot be solved by adding more instrumentation.

• •OpenTelemetry spans are observability, not compliance evidence under EU AI Act
• •Self-reported telemetry is structurally insufficient because the audited entity controls the record
• •Mutable storage, vendor control, and unserialized context create gaps no instrumentation can close

Generated with AI, which can make mistakes.