AI-Use Disclosure

1. How AI Is Used in This Product

The Perpetual Core RFP & Proposal Engine uses large-language models (LLMs) and embedding models throughout its core workflow. Specifically, AI is used to:

• Fit Scoring: Evaluate each funding opportunity against your organization profile and vault documents and assign a relevance score. Scores reflect AI judgment and should be treated as a starting signal, not a definitive assessment.
• Draft Generation: Produce full or section-level proposal drafts grounded in your uploaded vault documents (past proposals, annual reports, program descriptions) and the solicitation requirements.
• Adversarial Review:Critique drafts for logical gaps, unsupported claims, and weak sections — simulating a reviewer's perspective before you submit.
• Compliance Checking: Identify whether a draft addresses key eligibility and submission requirements stated in the solicitation, and flag missing elements.
• Embeddings & Semantic Search: Convert vault documents and opportunity text into vector representations to enable retrieval-augmented generation (RAG) — surfacing the most relevant portions of your vault when drafting a specific section.
• Rubric Extraction:AI extracts evaluation criteria from uploaded solicitation documents (Section L/M for federal RFPs; funder priority sections for grants) and structures them as scored criteria for use in the adversarial review pass.

2. Which AI Providers Process Your Data

When you use AI features in the Service, your vault documents, organization profile context, and opportunity details are sent to one or more of the following third-party AI providers:

• Anthropic (Claude): Used for proposal drafting, adversarial review, compliance checking, and fit-score narratives. Vault document excerpts and opportunity text are included in the prompt context. Subject to Anthropic's Privacy Policy.
• OpenAI:Used for embedding generation (text converted to vectors for semantic search). Input text is sent to OpenAI's embeddings API. Subject to OpenAI's Privacy Policy.

Under standard enterprise API terms, these providers do not use your input data to train their models. You should review each provider's terms to confirm this for your specific use case, particularly if your vault documents contain sensitive or proprietary information.

3. Human Review Requirement

AI-generated drafts are a starting point. The Service is designed to get you to a submission-ready state; it does not auto-submit anything to any government agency, foundation, or other funder.

Before submitting any proposal generated or assisted by this Service, you must:

• Review every section for factual accuracy, completeness, and compliance with the solicitation requirements.
• Verify that all data, statistics, certifications, and representations are accurate and authorized by your organization.
• Ensure compliance with any funder-specific policies regarding AI-assisted content, including any disclosure requirements (see Section 4).

The user is solely responsible for the accuracy, compliance, and legal sufficiency of any submitted proposal, regardless of how the draft was generated.

4. Federal & Grant Compliance Context

Organizations using AI assistance in responses to federal solicitations should be aware of applicable disclosure and compliance requirements. Requirements vary by agency, solicitation type, and program. The following is provided as general context, not legal advice; consult qualified counsel for your specific situation.

4.1 Federal Acquisition / GSA GSAR Clause 552.239-7001

Federal contractors and prospective contractors responding to General Services Administration (GSA) and other federal solicitations may be subject to GSA GSAR clause 552.239-7001 (Prohibition on Use of Artificial Intelligence for Certain Systems) and related AI-governance provisions. This clause governs the use of AI in IT systems that process federal data and may impose notification, transparency, or prohibition requirements depending on the solicitation and contract type.

If you are responding to a federal solicitation, review the full text of any AI-related clauses incorporated into that solicitation and determine whether disclosing the AI-assisted nature of your proposal content is required or appropriate. Perpetual Core LLC provides a draft AI-use disclosure notice within the compliance gate feature (Phase 19 of the product roadmap) to help you satisfy this requirement.

4.2 NIH Grant Applications

The National Institutes of Health (NIH) has issued guidance on the use of AI in grant applications. As of 2024, NIH generally expects applicants to disclose when AI tools were used to assist in writing an application if the funder's policies so require, and reviewers are expected to flag applications that appear to rely heavily on AI-generated content without appropriate human oversight. NIH policy evolves; always consult the current NIH policy and the specific Funding Opportunity Announcement (FOA) before submission.

4.3 Other Funders

Many other federal agencies, private foundations, and municipal funders are developing or have issued policies on AI use in grant applications. Review the applicable guidance for each funder and solicitation before using AI-generated content in a submission.

5. Controlling AI Use

You control when AI features are invoked. Specifically:

• AI drafting and scoring are initiated by explicit user actions (clicking "Generate Draft," "Score," etc.), not automatically.
• You may edit or entirely replace AI-generated content before submission.
• Vault documents are only sent to AI providers when you actively request a feature that requires them (e.g., generating a draft or running a vault search).
• If you need to limit AI processing due to policy constraints, contact us at legal@perpetualcore.com to discuss your options.

6. Data Retention for AI-Processed Content

Vault documents and proposal context sent to AI providers are processed in real time for each request. Under standard enterprise API terms, Anthropic and OpenAI do not retain this input data beyond the immediate processing window and do not use it to train their models. Perpetual Core LLC stores the outputs(draft text, embeddings, scores) in your account's tenant-isolated database until you delete them or close your account.

For detailed retention periods, see our Privacy Policy.

7. Contact

Questions about AI use in this product may be directed to: