Responsible AI

Last updated: June 29, 2026

HelmIQ uses AI to help bankers draft, summarize, and find things faster. This statement explains which providers we use, how your data is handled, and the guardrails we put around AI output. It is meant to be read alongside our Privacy Policy and Security & Trust Center.

1. Providers and routing

Model inference is routed through a single AI gateway (Concentrate), which forwards requests to downstream providers (such as Anthropic and OpenAI) under an enforced Zero Data Retention (ZDR) configuration. Audio transcription uses a no-retention endpoint. The current AI providers are listed on our subprocessors page.

2. No training on your data

No AI provider we use trains its models on your data. Our routing rejects providers that do not certify zero retention and no training, rather than silently falling back to a non-compliant path. This control is published as HELM-AI-01 on our Security page.

3. Human oversight

AI output in HelmIQ is positioned as a draft for a person to review, edit, and approve. Drafted emails, suggested tasks, meeting invites, and summaries are not sent or acted on automatically without a user in the loop. AI output is probabilistic and can be inaccurate, so you remain responsible for what you send and record.

4. Data minimization

We send the AI only the context needed for the task at hand, scoped to your organization. AI features never read across organizations, and the same tenant-isolation rules that govern the rest of the product apply to AI requests.

5. Your controls

Firms can limit or disable AI-assisted features. Connected mailbox, calendar, and file integrations are authorized by you and can be disconnected at any time, which stops further AI processing of that source. See the Privacy Policy for data access, export, and deletion rights.

6. Contact

Questions about how we use AI, or requests for our AI controls documentation for a vendor review: jack@helmiq.net.