AI Data Notice

Please read this before running live AI analyses on real architecture requests.

AI output is advisory

ARB Copilot uses AI to generate analyses, risk flags, reviewer routing suggestions, and recommendations. This output is advisory. The human decision owner — you, the architect, the ARB — remains accountable for the final architectural decision.

What to avoid sending

During early access, please do not enter:

  • Highly confidential or commercially sensitive deal information.
  • Regulated data (PHI, PCI, financial account data, controlled tech).
  • Personal data of identifiable individuals.
  • Production secrets, credentials, or proprietary source code.

Unless this is explicitly covered by a separate written agreement with us.

How AI processing works

Live AI analyses are sent from ARB Copilot's servers through its configured AI processing gateway to a third-party model provider. Requests are made over encrypted connections and are scoped to the submitting workspace. Alternative and private-VPC providers are on the roadmap for enterprise deployments. Public demo pages use static sample data only — no live AI is called. Retention and training behaviour depend on the active model provider; if your governance programme requires specific guarantees, contact us and we will confirm the current configuration. Current provider details are available on request.

ARB Copilot does not use tenant content to train its own models — we do not train, fine-tune or build models. Content you submit for a live analysis is sent to the configured AI subprocessors for inference only. What those providers retain, and whether they use content for their own model training, is governed by the applicable provider terms and configuration — we do not claim zero retention or a no-training guarantee on their behalf. Data residency for the model layer is not contractually fixed during the pilot. This is why regulated and sensitive data remains prohibited, and why AI output is advisory and always human-reviewed.

Recommended pattern for early proofs of value

Use synthetic or sanitized versions of your real architecture requests for the first few analyses. Replace customer names, product names, and identifiable details with placeholders. This lets you evaluate the product end-to-end with no data risk.

Mock / demo mode

Demo mode does not call any AI provider. It returns fixture analyses generated locally and is safe to use for screenshots, walkthroughs, and exploration.

Operator

ARB Copilot is operated by Hari Krishna Bodapati, based in India.

Contact

Questions about AI data handling: hari@eabyea.com.

Questions? Contact hari@eabyea.com. Last updated: June 2026.