AI governance

AI inside explicit boundaries.

What is in place today, where people stay responsible, and what we do not claim.

What is in place today

Consent before covered AI features

Signed-in staff must accept the current AI usage policy before AI requests on covered routes, such as assistant, compliance review and generation requests, are processed.

Redaction before a provider sees a request

The central AI call path removes common personal identifiers from prompts and from responses. Brokerages can add their own redaction patterns.

Provider controls per brokerage

Broker owners can block specific AI providers. A blocked provider is denied and the denial is logged. Providers can also be marked no-training, which attaches a no-training request to the call.

Payload capture and retention

Brokerages can turn off payload capture entirely or keep redacted payloads for 0, 7, 30 or 90 days. The default is 0 days.

A record of AI calls

AI calls are logged with provider, model, feature, token counts and whether a call was blocked. Where payloads are retained, the redacted text is stored encrypted.

Your own provider keys

Brokerages can connect their own provider credentials per provider, so requests run on their own provider account.

Human review

  • AI output is a draft. Qualified people remain responsible for valuations, legal notices, contracts, investment recommendations, payments and client communications.
  • An AI explanation is never an approval. Approval comes from the person accountable for the decision.

What we do not claim

  • This page makes no claim of any AI, security or privacy certification, and no independent audit is implied.
  • A data-residency preference can be saved, but it is not enforced in routing today, so we do not promise that requests stay in a region.
  • A no-training flag is a request to the provider. Provider retention and training terms are set by each provider and its agreement.
  • Redaction works on recognizable patterns and can miss information. Do not paste secrets or unnecessary personal details into AI features.
  • Trexia is not an authorized reseller of, or endorsed by, any model provider unless we say so in a specific written agreement.
AI usage policyTrust and securityPrivacy and security requests