Get Started

Choose your path

For Sellers

Considering selling?

See valuation, readiness, and 90-day transition expectations.

Valuation Calculator Fit & Alignment Learn How We Invest

Advisors

Represent a client or partner?

Submit a Teaser, See Process, Review SLAs.

See Verifications & SLAs Submit a Teaser

Investors

Evaluating Goldmont?

See Strategy, Review Governance, Request LP Overview.

Request LP Overview Investment Thesis

Resources

Want to self-serve?

Access research, templates, glossary & FAQs.

Research Templates FAQs

AI Provenance & Governance

How Goldmont Uses AI — Transparently & Responsibly

Owner: Compliance & Engineering • Last updated:

Scope & Purpose

We use AI to speed research, improve document quality, and enhance security reviews. AI does not replace human judgment for valuations, legal positions, or investment decisions.

Applies to public website content and non-binding communications.
Binding decisions (IC, legal, finance) follow human-owned processes.

Where AI Is / Isn’t Used

Used for

Drafting/editing internal docs (human review required)
Summarizing public materials & meeting notes
Detecting redline inconsistencies and formatting issues
Classifying inbound emails for routing
Illustrative graphics on public pages (AI-assisted graphic label shown)

Not used for

Final valuation or Investment Committee decisions
Authoring or approving binding agreements
Unlabeled synthetic testimonials or deal claims

Data Boundaries

Sources: public web, licensed vendor data, and our own docs with RBAC.
Sensitive data: no seller confidentials to external AI without executed NDA + written consent; only via approved vendors under DPA.
Retention: enterprise settings — no training on our data; zero or ≤ 30-day retention.

Models & Vendor Provenance

Enterprise LLMs (OpenAI, Anthropic, Google, Microsoft) are accessed via isolated accounts with audit logging; approved open-source models run inside our controlled environment.

Attribution: public artifacts carry an AI-assisted • Human-reviewed label naming the provider/model family where appropriate.
Versioning: model families and safety settings are reviewed quarterly.

Labels & Watermarking

Visuals: caption includes AI-assisted graphic; originals archived.
Text: “AI-assisted • Human-reviewed” pill + reviewer initials/date on posts where applicable.
Files: where supported, PDFs include metadata X-AI-Assisted: true.

Human-in-the-Loop Controls

Two-person rule on public claims, term sheets, and policy updates.
Red-team review for compliance-sensitive language (KYC/AML, legal).
Kill-switch: any team member can pause AI features pending review.

Safety, Fairness & Claims

No generation of protected-personal data or sensitive attributes.
No hallucinated credentials, partners, or closings.
Screening prompts are audited for bias; AI never auto-rejects counterparties.

Evaluation & Auditing

Quality gates: accuracy spot-checks on 10% sample; zero-tolerance for fabricated claims.
Logs: prompts/outputs logged with 90-day retention; access via RBAC.
Quarterly review: tool inventory, model versions, DPIA updates posted here.

Incidents & Requests

Report issues to compliance@goldmontholdings.com. We acknowledge within ≤ 1 business day.

Takedowns/corrections: public content corrected within 48 hours of validation.
Breach protocol: follows Security & Privacy Policy; vendor notification clauses enforced.

Policy Updates

We may update this page from time to time. Changes will be reflected here with an updated effective date and summarized in the changelog below.

2025-12-18: Initial publication.

Contact

Questions about this page may be directed to info@goldmontholdings.com.