Transparency Report

Don't Trust Us.
Verify Us.

In the age of AGI, "trust me" is not enough. We provide the logs, the audits, and the thought traces required for you to verify our claims.

Live Safety Telemetry

SYSTEM NORMAL

Total Requests (24h)

14.2M

Safety Interventions

0.04%

Below 0.1% threshold

Refusal Rate

1.2%

Avg. Trace Latency

42ms

External Audits

Oct 2025

ARC Evals: Model Autonomy Assessment

Independent assessment by the Alignment Research Center evaluating our Alpha-1 model for autonomous replication capabilities.

Sep 2025

System Card: Metanthropic-V1

Comprehensive system card detailing training data composition, safety benchmarks, and known limitations of our flagship reasoning model.

How to Audit a Thought Trace?

Every API response includes a `x-metanthropic-trace-id`. You can plug this ID into our Inspector Tool to replay the model's step-by-step reasoning process.