> For the complete documentation index, see [llms.txt](https://docs.autoenhance.ai/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.autoenhance.ai/ai-transparency.md). # AI Transparency Autoenhance uses AI to enhance and restage property photos. New rules in the EU and California increasingly require AI-edited images to be disclosed as AI content — usually as machine-readable metadata inside the file — so it's clear the picture has been altered by AI. To make this effortless, **every AI-processed image we produce is automatically labelled and documented** — you don't have to do anything extra. This page explains, in plain terms, what we disclose, how your copyright is protected, and how anyone can verify it. Where it helps, we include the technical specifics your developers or compliance team may need. {% hint style="info" %} This page explains how our features support transparency requirements. It is **not legal advice** — confirm your own obligations with your legal or compliance team. {% endhint %} ## Why this matters Two laws are driving this, with similar rules emerging elsewhere: * **The EU AI Act (Article 50).** From August 2026, images that are AI-generated or AI-manipulated must be marked in a **machine-readable** way, and people must be able to tell the content was AI-altered. * **California (AB 723).** AI-generated or AI-altered images must carry **machine-readable metadata** identifying them as AI content — a hidden disclosure embedded in the file, rather than a visible label. Both laws are satisfied by the hidden, machine-readable disclosure we embed in every AI export. We also offer an optional visible label for cases where you want an at-a-glance disclosure on the image itself. ## The three layers of disclosure We disclose AI use in three complementary ways: | Layer | Visible to a person? | On by default? | Best for | | ---------------------------------- | ------------------------- | -------------- | --------------------------------------------- | | **1. Visible "AI Modified" label** | Yes | Optional | An at-a-glance disclosure on the image itself | | **2. Embedded metadata** | No (readable by software) | Yes | Machine-readable disclosure inside the file | | **3. Content Credentials (C2PA)** | No (verifiable) | Yes | Tamper-evident, cryptographic proof of origin | ### 1. Visible "AI Modified" label A small **"AI MODIFIED"** badge (the EU-style disclosure icon) placed in the corner of the image. It's the most obvious, human-readable form of disclosure. Turn it on by adding `?visible_disclosure=true` when you download an enhanced image: ``` GET https://api.autoenhance.ai/v3/images/{image_id}/enhanced?visible_disclosure=true ``` ### 2. Embedded metadata We write standard photo-metadata tags into every AI-enhanced or restaged export, stating that the image was AI-modified by Autoenhance. This is **invisible** to someone looking at the photo, but any metadata tool can read it, and it uses the same industry-standard fields recognised by Adobe, Google and news agencies. The human-readable summary reads like this: ``` AI Modified Image by Autoenhance.ai using Autoenhance v5.0 on 2026-07-06. ``` For restaged images (where AI adds elements such as a new sky or furniture), it reads: ``` AI Modified Image with AI-generated elements by Autoenhance.ai using Autoenhance v5.0 on 2026-07-06. ``` {% hint style="info" %} We write the same disclosure across **EXIF, IPTC and XMP** so it survives no matter which of these standards a given piece of software happens to read. {% endhint %} ### 3. Content Credentials (C2PA) We also embed **Content Credentials** — a cryptographically signed record following the [C2PA](https://c2pa.org) standard backed by Adobe, Microsoft, the BBC, camera manufacturers and others. Where plain metadata can be edited or stripped, Content Credentials are **tamper-evident**: the record is sealed to the exact image, so any later change breaks the seal and a verifier will flag it. The credential records that Autoenhance produced the image, what was done to it (an AI edit, and any added elements), and the original photo it started from. {% hint style="info" %} Content Credentials signing is rolling out across exports. The embedded metadata (layer 2) and the visible label (layer 1) are available today. {% endhint %} ## What we disclose Everything we embed, in plain terms and with the exact technical location for your developers: | What we disclose | Plain meaning | Where it lives (technical) | | ----------------------------- | ------------------------------------------------------------------------------- | ---------------------------------------------------------------------------------------------------------- | | **AI system & version** | Which AI product and version edited the image (e.g. Autoenhance v5.0) | `EXIF:Software`, `XMP:CreatorTool` = `Autoenhance/`; also in the caption | | **AI modification statement** | A sentence saying the image was AI-modified, and whether AI elements were added | `EXIF:UserComment`, `IPTC:Caption-Abstract`, `XMP-dc:description` | | **Digital source type** | An industry-standard code marking it as a mix of a real photo and AI | `XMP-iptcExt:DigitalSourceType` = `compositeWithTrainedAlgorithmicMedia` | | **Date processed** | When we produced the image | `EXIF:ModifyDate`, `XMP:ModifyDate` | | **Content Credentials** | Signed proof of who made it, what was done, and the source photo | C2PA manifest (`c2pa.opened` / `c2pa.edited` / `c2pa.placed` actions, original as an `inputTo` ingredient) | | **No-AI-training notice** | A request that the image not be used to train generative AI | `XMP-plus:DataMining` = `Prohibited`; C2PA training/data-mining set to *not allowed* | | **Asset reference** | An opaque ID we can trace back to the source image | `XMP-iptcExt:DigitalImageGUID` | ## Your copyright and credits We are careful to protect your rights, not claim them: * **We never overwrite your copyright.** If your original upload already has a caption, author/byline or copyright notice, we **preserve it** and add our AI disclosure alongside — we don't replace it. * **You are credited as the author.** In the Content Credentials, Autoenhance is recorded only as the *editing tool*. Your authorship and copyright (read from your original file) are carried through as the author of the photo. * **We add an AI-training opt-out** so your images are marked as not permitted for generative-AI training or data-mining. * **We protect privacy.** GPS and location data from the original file is deliberately **removed** from exports, so property locations aren't leaked. ## How to verify Anyone — you, your client, or a regulator — can check the disclosures: * **Content Credentials (recommended).** Open the image with a Content Credentials inspector such as the Content Authenticity Initiative's [Verify tool](https://contentcredentials.org/verify) or Adobe's Content Credentials. It shows that Autoenhance produced the image, what was done, and whether it has been altered since. * **Metadata viewers.** Any metadata tool reveals the embedded tags. For developers, [ExifTool](https://exiftool.org) is the quickest check: ```bash exiftool -G1 enhanced.jpg # Look for: # Software : Autoenhance/5.0 # DigitalSourceType : https://cv.iptc.org/newscodes/digitalsourcetype/compositeWithTrainedAlgorithmicMedia # UserComment : AI Modified Image by Autoenhance.ai ... # DataMining : Prohibited ``` ## For developers and compliance teams A few implementation details worth knowing: * **Standards.** We follow [C2PA / Content Credentials](https://c2pa.org), and the [IPTC Photo Metadata](https://www.iptc.org/std/photometadata/specification/IPTC-PhotoMetadata) `digitalSourceType` controlled vocabulary. The disclosure is written across **EXIF, IPTC and XMP** for maximum compatibility. * **Formats.** Disclosures are embedded in **JPEG, PNG, WebP, AVIF and JPEG XL** exports. * **Only AI outputs are labelled.** Disclosure is applied when you download an AI-produced image (enhanced or restaged). Downloading your **original, unedited** image is never labelled as AI-altered. * **Provenance chain.** The original photo is recorded as an input to the edit (an `inputTo` ingredient), so the credential links the original capture to the enhanced result. * **Applied at export, always current.** Disclosures are added when an image is downloaded, so improvements to our disclosure automatically apply to new and existing images the next time they're fetched. ## Quick checklist * **Visible label?** Add `?visible_disclosure=true` when downloading. * **Hidden disclosure + Content Credentials?** Automatic on every AI output. * **Verify?** Use [contentcredentials.org/verify](https://contentcredentials.org/verify) or `exiftool`. * **Your copyright?** Preserved, you're credited as author, and marked not for AI training. --- # Agent Instructions This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com. ## Querying This Documentation If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question. Perform an HTTP GET request on the current page URL with the `ask` query parameter, and the optional `goal` query parameter: ``` GET https://docs.autoenhance.ai/ai-transparency.md?ask=&goal= ``` `ask` is the immediate question: it should be specific, self-contained, and written in natural language. `goal` is optional and describes the broader end goal you are ultimately trying to accomplish on behalf of the user. GitBook uses it to tailor the answer towards what is most useful for that goal. The response will contain a direct answer to the question and relevant excerpts and sources from the documentation. Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.