AIM Transparency – AI Disclosure for EU AI Act

Registro de alterações

2.0.2

• Fix: Algorithmic can now be set in bulk. Six of the seven source types were in the Media Library bulk menu. Algorithmic, for procedural work like fractal or formula art, was left out, so the only way to set it was one image at a time. It sits with the others now, translated in all 21 languages.

2.0.1

• Improvement: the Media screen leads with your library. The eight source-type explanations used to sit above it, so the actual work started below the fold. They now sit under the list and fold away, and whether you leave them open or closed is remembered. A link at the top jumps to them when you want them.
• New: filter your library by With AI or Without AI. The filter offered only “everything recorded” before, which could not answer the obvious question of which images are the AI ones.
• Improvement: the screen you were on survives a reload. Refreshing used to drop you back on Overview. The tab is now part of the address, so a reload keeps your place and you can link someone straight to a screen.
• Fix: AVIF and TIFF now appear on the Metadata screen. The plugin has always written them wherever your host has exiftool, but the screen only knew four formats, so those images were missing from it entirely. On a server without exiftool they now say so plainly instead of showing an unexplained zero.
• Fix: AI (basic) had the same icon as AI Generated, so the two were indistinguishable in a list until you read the label. It has its own now.
• Improvement: clearer wording on the Shortcodes screen, and the picture beside a disclosure is now called a symbol throughout, so it is no longer confused with the mark written inside your image files.

2.0.0

• New: the plugin now speaks 22 official EU languages: Bulgarian, Croatian, Czech, Danish, Dutch, English, Estonian, Finnish, French, German, Greek, Hungarian, Italian, Latvian, Lithuanian, Polish, Portuguese, Romanian, Slovak, Slovenian, Spanish and Swedish. Every visitor-facing disclosure and the whole admin dashboard are translated, with the legal wording taken from each language’s own official text of Regulation (EU) 2024/1689 rather than translated from English. Irish and Maltese are the two official EU languages WordPress itself does not ship. A disclosure only informs someone if it is in a language they read.
• New: the dashboard now looks like WordPress. A compliance tool should read as part of your admin rather than as a guest that brought its own furniture, so the default is a new WordPress-native layout built from the same container, heading and tab strip that every core settings screen uses. It picks up your admin colour scheme on its own, and there is nothing of ours to drift out of date when WordPress changes. The Neo and Classic themes are still there, and if you have already chosen one, updating does not change it under you.
• New: the Overview screen now fixes things instead of only reporting them. Every finding carries the action that resolves it, so an image missing its mark, a chat widget nobody has disclosed, or an unrecorded Article 4 statement is one click from done rather than a trip to another tab. Findings are sorted into what needs attention, what is already working and what is genuinely optional, so a clear site reads as clear instead of as a wall of warnings. Each change is dated in a Recent changes list with an Undo beside it, so nothing the screen does is one-way.
• Improvement: the dashboard reads WordPress’s own translation runtime instead of shipping a second copy of it, so the admin screens stay in step with the language files the rest of the plugin loads.
• New: multilingual sites are handled properly. Tested with Polylang: a source type set on an image is carried to that image’s translated copies, so the badge no longer disappears on a translated page when media translation is switched on. The plugin ships a wpml-config.xml that both Polylang and WPML read. WPML itself is built for but not yet tested, and the readme says so rather than claiming otherwise.
• New: the dashboard asks for corrections in any language other than English. The translations were made with AI assistance and checked against the Regulation’s own text in each language, but not by a native speaker of every one, so the plugin says so where it matters and gives you an address to report anything that reads wrong.
• Fix: the plugin no longer assumes every chat tool is an AI. It used to treat all 304 chat systems it knows about as AI assistants, including Tawk.to, Olark, Intercom and Crisp, which are live chat answered by real people. Following that advice would have told your visitors they were talking to an AI while a colleague was typing. Every chat tool is now classified by who actually replies, and the plugin says one of three things: this one replies with AI, this one can be either depending on your setup, or this one is answered by your team and Article 50(1) does not apply.
• Improvement: detection reads more of your site (your shop and product pages as well as the home page, posts and pages) and keeps only the part of each page it needs, so the check covers more ground while storing about a tenth as much.
• Improvement: a new aicl_scan_urls filter lets you point detection at any page a widget only appears on, such as a checkout or a members area.

1.3.1

• Fix: “We could not check your pages” appeared on sites that were working perfectly well. Detection reads a few of your own pages to find chat widgets, and a server cannot always reach itself at its public web address: local installs publish on a different port than the web server listens on, an origin behind a proxy often cannot resolve its own hostname, and some firewalls block a site from requesting itself. The scan now falls back to asking the server directly, on its own address, so it succeeds in these setups instead of reporting a failure. Sites that were already fine are unaffected and make no extra requests.

1.3.0

• New: the plugin speaks German, French and Spanish. Every visitor-facing disclosure is translated: the badge wording, the “this image was generated by an AI model” explanation, the disclosure modal and the chatbot notice, plus the admin menu, the media-library labels and the Article 4 readiness checklist. A disclosure only informs someone if it is in a language they read, so this is the release that makes Article 50 work outside English. The dashboard screens themselves are still English and are next. (Bundled translations were previously never loaded at all, which is fixed here.)
• New: detection now finds AI chat widgets that were not installed as WordPress plugins. Alongside plugins and script addresses, it recognises the JavaScript a widget leaves in the browser and the cookies it sets, so a chat tool pasted into your theme, added through a tag manager or served from your own CDN is still found. Same 672 systems, more ways of spotting them. Matching is deliberately strict: a widget’s name mentioned in a blog post, quoted in an install tutorial, or sitting in a file name is not treated as that widget running.
• New: detection reads three pages instead of one: your home page plus your most recently updated post and page. A widget that only loads on posts is no longer missed because the home page happens not to show it.
• Improvement: the dashboard now admits when it could not check. If your pages cannot be read (a firewall blocking loopback requests, a password-protected site, a server error), you get a plain explanation and what was still checked, instead of a confident “You’re all clear”. When it is genuinely clear, it now tells you what it covered. A scan that could not run is no longer reported as a scan that found nothing.
• Security: the CSV compliance export now neutralises spreadsheet formula injection. A filename or caption beginning with =, +, – or @ was previously written to the CSV as-is, where Excel and Sheets would treat it as a formula. Exports also carry a UTF-8 marker so accented filenames stop mangling.
• Fix: the self-scan no longer skips TLS verification.

1.2.9

• Fix: on a WooCommerce variable product, choosing a variation swapped the picture but left the previous image’s disclosure on it. A photograph could end up wearing an “AI Generated” badge, and an AI variation could end up with none. The badge now follows the image: WooCommerce publishes each variation, so the plugin adds its verdict there and the browser re-decides whenever the picture changes.
• Fix: the dashboard under-reported large libraries. Figures were gathered from a bounded page of attachments, so a site past that bound saw a smaller flagged count than it really had, and disclosure coverage was calculated over the truncated set. Totals are now aggregated in the database and are correct at any size. If your numbers jump after updating, the new ones are the true ones.

1.2.8

• Fix: the official EU mark on a disclosure notice rendered squashed. The Commission’s set has a square mark and two wider lockups that already spell out their own label; the notice now always uses the square one, undistorted, since the sentence beside it is already saying the same thing.

1.2.7

• New: a Shortcodes panel in the dashboard, under Disclosure. Pick the type, the style, your own wording, colours and mark, watch it render, and copy the finished shortcode. The shortcodes were previously documented but invisible in the plugin.
• New: [aicl_disclosure] takes bg and color, so a notice can match your theme instead of always being the default near-black pill. A pair below the WCAG AA minimum of 4.5:1 is rejected and the default renders instead, because a disclosure nobody can read is not a disclosure.
• New: mark chooses the glyph: aim (the built-in mark), eu (the official European Commission AI-content icon for the matching source type), or none. A logo attribute places your own image beside the mark, never instead of it.
• Improvement: each source type has its own icon in the dashboard lists (camera, pen, layers, shapes) rather than one cog standing in for everything that is not AI.
• Fix: the toggle control centres its knob correctly on both themes.

1.2.6

• New: the dashboard has been redesigned. Every screen is now a single sheet divided by hairlines instead of a grid of separate cards, so Overview reads top to bottom: your readiness score, the figures behind it, what is running on your site, then your library. Readiness gained a numbered Article 4 checklist with proper evidence fields, Media shows each image’s disclosure state at a glance, and Upgrade sets out plainly what is free and what Pro adds.
• New: Detection now recognises 671 AI systems, up from a short list of image tools. It separates chatbots and assistants (Article 50(1)) from image and content generators (Article 50(2)), so you can see which duty applies to what.
• New: reusable disclosure shortcodes. [aicl_disclosure] places an AI notice anywhere shortcodes run, with a type (generated, edited, assisted or chatbot) and a style (inline, badge or banner). [aicl_ai_notice] covers the chatbot case.
• Improvement: Detection also reads your own front page, so AI widgets that load in the browser are found, not just plugins on disk. It stays on your server; nothing is sent to a third party.
• Improvement: uploading an image refreshes the Detection result straight away instead of waiting for the cache to expire.
• Improvement: clearer wording throughout the dashboard, and a refreshed getting-started screen in both the Neo and Classic themes.

1.2.5

• Improvement: the getting-started screen has fresh artwork that matches your Neo or Classic dashboard theme, in a cleaner full-bleed layout.
• Improvement: the Detection card is more reliable. Re-scanning clears resolved warnings right away, it respects images you have marked as human-made instead of re-flagging them, and it lists AI-capable plugins by their real names.
• Improvement: clearer, more natural wording across the dashboard and the getting-started screen.
• Fix: the dashboard now keeps your chosen theme (Neo or Classic) consistent everywhere, including the getting-started screen.

1.2.4

• New: a Detection card on the dashboard that surfaces AI you have not disclosed yet. It finds images in your library whose files read as AI-generated (a declared source type, or a known generator name like Midjourney, DALL·E or Firefly) but are not flagged, and lists active plugins that can generate AI images. Everything is computed locally, with no external requests.

1.2.3

• New: a one-screen welcome after activation that walks you to your first flagged image: pick an image, choose how it was made, and see it disclosed straight away. Reopen it any time from Transparency > Getting started.
• Maintenance: minor internal hardening and housekeeping.

1.2.2

• Improvement: the Pro library scan now runs to completion from one click: the dashboard processes the library in batches with a live progress bar (inspected / adopted / remaining) and a cancel button, instead of stopping after one pass.
• Improvement: the scan result now reports what discovery found (images adopted from provenance already declared inside the files), alongside the strip-guard counts.
• i18n: the disclosure modal footer, its close button and the badge screen-reader fallback are now translatable.

1.2.1

• New: human-provenance source types: flag images as Camera Photo (digitalCapture), Human Created (digitalCreation, e.g. 3D renders or digital art made without generative AI) or Composite (composite, a non-AI mix such as photos + renders). Each embeds the current IPTC Digital Source Type into the file and the page JSON-LD, with no visible badge by default: honest provenance for human-made work, not just AI labels. (An optional visible badge for these types is available with the Custom badge style.)
• Improvement: embedding is now idempotent: files that already carry the exact Digital Source Type value (for example C2PA-signed camera captures) are left byte-for-byte untouched.
• Improvement: source-type names and the visitor-facing disclosure texts are now translatable.
• Fix: images whose badge is hidden (e.g. the Algorithmic type) now correctly emit their machine-readable JSON-LD, as the settings always described.
• Accessibility: custom-style badges on non-AI types no longer announce “artificially generated” to screen readers.

1.2.0

• New: a fresh default dashboard style (“Neo”), with a one-click toggle back to the original monochrome look, in Settings > Appearance or the palette button in the top bar. Both styles support light and dark mode. Fonts are bundled locally (Inter + Bricolage Grotesque, SIL OFL); the plugin still makes no external requests.
• New: the badge live preview now demonstrates the badge on bundled sample images (our own AI-generated artwork, labelled as such), with a small slider to switch scenes. Docs and Support links added to the dashboard sidebar.

1.1.0

• New: AI-assistant (chatbot) disclosure for EU AI Act Article 50(1): tell visitors when they are interacting with an AI rather than a human. Drop the [aicl_ai_notice] shortcode into any page with a chat UI, or enable the optional floating notice that appears when a known chat widget is detected. Configurable disclosure text and position.

1.0.0

• Initial release: media flag, bulk actions, visible badge across Enfold/standard/block themes, optional universal badge coverage, IPTC/XMP embedding (exiftool + pure-PHP JPEG/PNG), JSON-LD, free CSV compliance export, React dashboard, opt-in transparency directory.