AI TRANSPARENCY NOTICE

Application: Seaworthy (za.co.digitalfreedom.seaworthy)

How Seaworthy uses artificial intelligence — and what that means for you

Effective Date: 5 October 2026

Provider:
DigitalFreedom
A brand of DigitalFreedom Global LLC
30 N Gould St, Ste N
Sheridan, WY 82801
United States
Contact: mailto:hello@digitalfreedom.co.za
Data protection: mailto:data-protection@digitalfreedom.co.za
Website: https://digitalfreedom.co.za

0. SCOPE AND PURPOSE

This Notice describes — in plain language — how Seaworthy uses artificial-intelligence (AI) systems, what that means for your data, and how to opt out where applicable.

It is designed to satisfy:

Transparency obligations under Art. 13 / 14 GDPR
Article 50 of the EU AI Act (Regulation (EU) 2024/1689) — transparency duties for providers and deployers of certain AI systems, applicable since 2 August 2026 (Art. 113(b)). Art. 50 is not the general-purpose-AI provision: the GPAI obligations sit in Chapter V, Art. 53 to 55.
Art. 13(2)(f), 14(2)(g) and 15(1)(h) GDPR — meaningful information about the logic involved in automated decision-making. Art. 22 is a separate matter: it grants a right not to be subject to a solely automated decision with legal or similarly significant effect, which is not an information duty.
the OECD AI Principles and UNESCO Recommendation on the Ethics of AI as best-practice references
platform-store policies (Apple, Google) on AI feature disclosure

Seaworthy is distributed through the Apple App Store and TestFlight in its offered markets. This Notice applies globally. The Provider applies the EU AI Act framework as a baseline. Local AI-disclosure rules are assessed where they apply and are named only once that assessment has been made — a general claim of compliance with regimes such as the Chinese Interim AI Measures or Brazil’s PL 2338/23, a bill that is not in force, would assert more than has been examined. Where a market is deselected, its rules do not arise at all; see the market decisions in the product profile.

1. WHERE AI APPEARS IN Seaworthy

Feature	What the system does	Visible result
Vision text recognition (OCR)	Reads text in a selected or captured image; local parsing identifies dates, serial numbers and model numbers	Recognised text and proposed item fields for review
Vision image classification and SigLIP 2	Classifies equipment; SigLIP 2 compares the image with the bundled maritime vocabulary on-device	Proposed equipment title and category
Apple Foundation Models (Apple Intelligence)	On supported devices with Pro access, uses recognised text and category hints to suggest a title and maintenance interval	Editable suggestions in the item draft
Optional Qwen2-VL-2B-Instruct, 4-bit	On compatible devices with Pro access, recognises equipment from an image and generates draft details using the additional downloaded model	Editable Pro recognition results

These are suggestions, not verified equipment instructions. Review and correct every result before saving or relying on it.

2. WHO PROVIDES THE AI

All recognition and suggestion processing takes place on your device. There is no external inference processor and no remote inference API. Apple’s Vision and Foundation Models frameworks, the bundled SigLIP 2 model (Google’s open model, licensed under Apache-2.0) and the optional Qwen model execute locally. Their suppliers do not receive your inputs merely because those models run.

The optional Qwen2-VL-2B-Instruct 4-bit model is downloaded from the Publisher’s Civo object storage at objectstore.lon1.civo.com. It is about 1.2 GB; newly downloaded model weights are checked against a pinned SHA-256 digest before use. Civo provides the file download and receives ordinary network-request data, but does not run the model or receive your photos or prompts.

See the Open Source Notices for software and model licence terms, and the Privacy Policy for Apple’s separate role as the provider of optional private iCloud.

3. WHAT DATA GOES TO THE AI

The on-device pipeline processes the photo you selected or captured, recognised text, category hints and the language or settings needed to generate the draft. These inputs remain on the device during inference. No user content is sent to a model provider, Civo or an external inference API.

If you save the result, it becomes part of your logbook and can sync through your private iCloud if enabled. If you voluntarily send a result to support, the material you send leaves the device as part of that support request; neither operation is remote inference.

4. WHAT THE AI DOES WITH IT

The pipeline produces local classifications, text recognition and draft suggestions. SigLIP 2 compares the image with the bundled maritime vocabulary on-device. It does not train or fine-tune a provider’s model on your content, and there is no server-side inference retention period.

Saved item details and recognition results follow the logbook’s local and optional private-iCloud storage and deletion rules. Downloaded model files are cached on the device so routine later use needs no further model download. Switching back to the base recognition option stops use of the optional Pro model; it does not imply that the cached files have been erased.

5. LIMITS OF AI

You should know:

AI outputs can be wrong. Generative AI can produce inaccurate, biased, or misleading content. Verify before relying on AI outputs for decisions that matter.
AI is not a professional. AI is not a lawyer, doctor, financial advisor, or therapist. For decisions in those domains, consult a qualified professional.
AI does not understand context the way humans do. It can miss nuance, sarcasm, or specialised knowledge that is obvious to a human reader.
AI can produce content that resembles real people. We do not intend AI to impersonate; if you encounter output that does, report it via mailto:support@digitalfreedom.co.za.

6. NO AUTOMATED LEGAL OR SIMILARLY SIGNIFICANT DECISIONS

Seaworthy does not use AI to make decisions that produce legal effects on you or similarly significantly affect you (Art. 22 GDPR). Outputs are informational. Seaworthy has no user account or remote fair-use enforcement. StoreKit verifies Pro purchase entitlement; model output does not decide purchase or refund rights.

If a feature ever introduces automated decision-making with legal or similarly significant effect, we will:

update this Notice
obtain the appropriate lawful basis (typically explicit consent under Art. 22(2)(c) or contract necessity under Art. 22(2)(a))
provide meaningful information about the logic involved
guarantee a right to human review (Art. 22(3))

7. OPTING OUT

Every photo you add for a new item is analysed on your device. All suggestions are drafts you can edit or discard before saving.

In Profile, choose the base recognition option to stop using the optional downloaded Pro model. Base recognition still uses local Vision and SigLIP 2 and, where available and applicable, Apple Foundation Models. There is no global switch that disables all recognition. Local maintenance logging does not require the optional model download.

8. AI-GENERATED CONTENT INDICATION

Recognition results are proposed item details for you to review and edit. Seaworthy does not generate synthetic images, audio, video or deepfakes.

The current implementation does not attach a standard machine-readable provenance label to exported or saved suggested text. This is a limitation: an editable draft or visible feature description is not a claim that Art. 50(2) machine-readable marking has been implemented.

9. PROHIBITED INPUTS

Use only photos and other material that you are authorised to process. Do not use the features unlawfully or infringe third-party rights. Seaworthy performs local analysis and has no remote moderation, content-monitoring or inference-quota enforcement service.

10. YOUR RIGHTS

The rights and contact routes in the Privacy Policy apply. You can edit and delete your saved logbook records in the app. The Provider has no remote copy of your inference inputs or outputs to retrieve or delete. Apple’s controls govern your separate private-iCloud and backup copies. For material you voluntarily sent to support, contact mailto:data-protection@digitalfreedom.co.za.

You may complain to the data-protection supervisory authority of your habitual residence, place of work or the place of the alleged infringement (Art. 77 GDPR).

11. TRANSPARENCY ABOUT MODEL CAPABILITIES

The models and frameworks are identified in §§ 1–2 and the Open Source Notices. Vision and SigLIP 2 identify text and likely equipment categories; Apple Foundation Models and the optional Qwen model produce draft text and maintenance suggestions. Availability depends on device capabilities, platform support and Pro entitlement. Failed recognition can fall back to a partial draft or manual entry.

These features cannot establish seaworthiness, certify safety equipment or replace manufacturer instructions, inspection or qualified maintenance advice.

12. INCIDENT REPORTING

Report an incorrect or harmful suggestion through https://support.apps.digitalfreedom.co.za or mailto:support@digitalfreedom.co.za. Include the relevant result, date and device/app version, and only the input material you choose to share. Sending a report is voluntary and transmits its contents to support.

We review reports and may update prompts, parsing, vocabulary or model selection.

13. CHANGES TO THIS NOTICE

We may update this Notice from time to time. Material changes (new AI provider, new high-risk AI feature, change in retention or training policy of the AI provider) are notified at least 15 days before they take effect. The “Effective Date” at the top reflects the latest revision.

14. CONTACT

DigitalFreedom
A brand of DigitalFreedom Global LLC
30 N Gould St, Ste N
Sheridan, WY 82801
United States

AI questions: mailto:support@digitalfreedom.co.za
Data protection: mailto:data-protection@digitalfreedom.co.za
General: mailto:hello@digitalfreedom.co.za
Website: https://digitalfreedom.co.za