How we use AI.

Your reviewers score. CallSherpa's AI will not fill in a rubric, produce a score, rank a shortlist, or say “this is strong” or “this is weak.” Judgments of merit come from your reviewers.

One exception, which we would rather name than hide: eligibility screening. If you switch it on, AI checks a submission against the eligibility criteria you wrote and suggests pass / flag / fail with an explanation. That is a check against your rules, not a judgment of merit — and a human screener confirms every one before any applicant is filtered out (see 2).

Other platforms draw the line elsewhere. Submittable's Automated Review applies “automated scoring based on your own custom-defined rubric criteria,” with “human validation of a subset of applications.” OpenWater's AI will “score entries, add commentary, and identify areas needing attention for judges.” Both are reasonable answers to volume. Neither clears a policy that says AI may not evaluate an application.

No submission is rejected, shortlisted, or filtered by AI. AI may suggest pass / flag / fail during eligibility screening, with an explanation, but a human screener confirms the decision before any applicant is filtered out. Every submission that meets your eligibility criteria reaches a human reviewer.

Reviewers form their own assessments. AI drafts text (a summary, a translation, a decision letter) only when a reviewer directs it to. AI output never appears unprompted in a reviewer's queue.

We send submissions only to model providers that contractually do not train on customer data. We enforce this on every API call, including when an organizer brings their own OpenRouter key.

This page is the disclosure. It names every model we call by its exact id, says what each one is used for, and is checked against the code that does the calling on every build — so a model cannot reach the platform without appearing in the table below.

What we do not do today, since this page is only worth anything if it is exact: there is no per-item label inside the product. An applicant looking at their own submission cannot see that a summary or a translation was AI-drafted, or which model produced it. We would rather name that gap than imply a surface that does not exist. If we ship per-item labeling, this page changes before it does.

AI features are optional. You can run a call on CallSherpa with AI fully off. AI can also be toggled per feature (summaries, translations, decision letter drafting) independently.

CallSherpa uses a small, vetted set of models, every one of them reached through OpenRouter. Each row below names the exact model id our requests carry. This table is checked against the code that calls the models on every build, so a model cannot reach the platform without appearing here.

No data retention. On every OpenRouter request that supports it, CallSherpa sets provider.data_collection: "deny". OpenRouter then routes only to provider endpoints that do not log or train on inputs, and rejects the request rather than falling back to one that would. We send this on every such call, on every plan, including when an organizer brings their own OpenRouter key.

One kind of request doesn't carry that per-request setting: requests to the decision model (typesafe/jev-1.13, above) — reviewer chat routing and the screening-criteria check — go to a different OpenRouter endpoint that accepts no provider field. They are covered instead by the same guarantee at the account level — data retention is disabled on the CallSherpa OpenRouter account, which applies to every request we send, whether or not the individual request can carry the flag.

With BYOK, you add an OpenRouter API key to a call and that call's AI usage bills to your own OpenRouter account. CallSherpa still chooses which models are called. The platform routes through the same vetted no-training models we use on AI Included, listed above. BYOK changes who pays for usage. It does not change which models CallSherpa uses or how submission data is handled.

Free OpenRouter models that train on inputs are not used by CallSherpa on any plan.

The model list is reviewed quarterly and updated whenever a provider's policy changes.

• No AI decides anything about your submission. A human reviewer reads it and makes the decision.
• AI may be used to summarize or translate your submission, or to draft a response to you. A human directs it and a human decides. Today this is not labeled item-by-item in the product — the models that can do it, and what each is used for, are listed on this page.
• Your submission is not used to train AI models. It's sent only to providers that contractually agree not to train on it.
• You can ask the organizer which AI features were switched on for the call. Organizers choose that per call, and can run a call with AI fully off.

• You can turn AI off entirely. Per call, in your settings.
• The models CallSherpa can call, and what each is used for, are published on this page. Per-call AI spend reporting is not an organizer-facing surface today.
• You decide what AI helps your reviewers with. Summaries, translations, decision letter drafting. Toggle each one.
• You can adopt reviewer guidelines aligned with federal grant standards. NIH prohibits peer reviewers from using AI for critiques; NSF says current generative AI does not meet their confidentiality bar. CallSherpa's six commitments map cleanly onto a posture a federal-grant-aligned program can adopt.
• If you bring your own OpenRouter key, CallSherpa still picks the models. You pay for usage on your account; the platform routes through the same vetted no-training models we use on AI Included. This keeps commitment #4 intact even on BYOK.