AI Training & Model Usage
Last updated: May 17, 2026
1. Our commitment
CVJinny does not use your prompts, uploads, generated outputs, or any customer data to train, fine-tune, evaluate, red-team, or improve any AI model — ours or anyone else's. Customer data flows through AI providers only to produce the output you requested, and only under the provider's no-training enterprise/API terms.
2. Provider guarantees
| Provider | Purpose | No-training basis | Zero-retention? |
|---|---|---|---|
| Google Gemini API | Text + image generation | Gemini API Terms § "No training on your data" for paid tier | No persistent retention beyond abuse-monitoring window |
| Hugging Face Inference API | Auxiliary classifiers, moderation | HF Inference TOS — user inputs not used for training | Request/response logs purged per HF policy |
3. What upstream providers may do
Providers may retain inputs/outputs briefly (typically up to 30 days) for abuse detection and quality assurance. They process this data as independent controllers under their own terms; we have reviewed and documented each in the sub-processor list.
4. Your data & your outputs
- You own the inputs you provide.
- Subject to the Terms and applicable law, you own the AI outputs we deliver to you. Copyrightability of AI-generated works varies by jurisdiction — see our Disclaimer.
- We claim no license to use your outputs for training, marketing, or resale.
5. Synthetic-media disclosure
Applicable law (EU AI Act Art. 50, California AB 2655/2839, and similar statutes) increasingly requires visible disclosure when AI-generated content depicts real people or political subjects. Compliance is the user's responsibility. We surface a "Generated with AI" toggle in the composer and recommend leaving it on by default.
6. Do-not-train signals for third-party content
When you point our trend/competitor analysis tools at public web pages, we respect robots.txt, noai, noimageai, TDM Reservation (ai.txt), and the emerging C2PA do-not-train manifests. Content published by others under these signals will not be ingested.
7. Watermarking & provenance
Generated images include invisible watermarking where the upstream provider supports it (SynthID for Gemini) and optional C2PA manifests. Do not remove watermarks; see AUP.
8. Reservation of Rights — Providers, Models & Policy
The Owner reserves the absolute, exclusive, and unfettered discretion, with or without prior notice and without liability, to: (a) add, change, replace, suspend, or discontinue any AI provider, model, fine-tune, prompt template, moderation rule, or output-pipeline component; (b) modify, restructure, or replace this AI Training & Model Usage policy at any time with effect upon posting (subject only to legally-required minimum notice); (c) introduce or remove any feature, plan, integration, price, fee, FX rate, credit weight, quota, discount, coupon, reward, or facility — partially or fully; and (d) pass through any AI-provider, infrastructure, or third-party cost increase to your invoices or credit weights, immediately or with delay, with no obligation to pass through cost decreases — all pursuant to the Universal Reservation of Rights in Terms of Service section 1.1, the Authorized-Use License in section 1.2, the Benefits Reservation in section 1.3, the Dynamic Pricing & FX clause in section 1.4, and the Comprehensive Owner Protections in section 18, all of which are incorporated into this page in full. The exercise of any right under this section gives rise to no refund, credit-back, pro-rated reimbursement, alternative compensation, or service credit of any kind; the No-Refund Policy at /refund-policy applies in full and is incorporated by reference. The no-training commitment in section 1 is a contractual undertaking that the Owner can only narrow on a forward-looking basis with notice; it does not retroactively affect data already lawfully collected.
9. Authorized Use of Aggregated & Derived Data
For clarity, while the Owner does not use customer prompts, uploads, or generated outputs to train AI models, the Owner does retain the broad license to collect, derive, aggregate, anonymize, model, share, license, sell, and otherwise commercially exploit non-content telemetry, usage signals, billing behavior, and aggregated insights as set out in Privacy Policy section 4a and Terms of Service section 1.2. The Owner is the sole and exclusive beneficiary of any revenue derived therefrom.
10. Contact
Questions about AI data handling: [email protected]