Copyright.

Signals respected · gateway-side detection

1. robots.txt

   Standard web-crawler robots.txt directives respected. If your site disallows BRMSTE substrate crawlers, the gateway refuses to ingest that source.

2. ai.txt

   The emerging ai.txt convention (AI-specific crawler directives) is honoured. Includes opt-out, opt-in, and conditional-use signals.

3. C2PA do-not-train

   C2PA Content Credentials manifest with do_not_train=true assertion. When attached to image/audio/video/text content, the substrate refuses to ingest the asset.

4. TDM Reservation Protocol (IETF draft)

   Future-compatible: the emerging IETF TDM Reservation Protocol headers will be honoured the day they are ratified.

What BRMSTE commits to

1. 01

   TDM opt-out detection

   The BRMSTE gateway at brm/gateway/dispatch.py implements state-of-the-art detection of TDM opt-out signals. Failure to honour an opt-out triggers the ATMA-SAMARPAN + BALIDAN two-stage sequence at the gateway — handover (cache surrendered · refusal recorded · state transferred) then end (call refused · class=termination StageRecord anchored).

   Art. 53(1)(c)
2. 02

   Training-data summary

   Every BRMSTE-trained model (including the sub-3B distillations under continuation #54 BRMSTE-LEANQLORA-SB2026) publishes a training-data summary per the AI Office template. The summary cites source categories, licensing status, opt-out detection method, language/domain distribution.

   Art. 53(1)(d)
3. 03

   Operator override prohibition

   Operators cannot override the TDM opt-out detection. The substrate's response to an opt-out signal is non-negotiable at every tier — including industrial and sovereign deployments.

   BRMSTE +
4. 04

   Citation discipline

   Every substrate output that draws from a primary source carries the citation in the output (footnoted, manifest-embedded, or inline). The GI tier specifically produces Wikimedia-grounded dossiers with explicit citation per claim.

   BRMSTE +