The IPTC has responded to a multi-stakeholder consultation on the recently-agreed European Union Artificial Intelligence Act (EU AI Act).

Although the IPTC is officially based in the UK, many of our members and staff operate from the European Union, and of course all of our members’ content is available in the EU, so it is very important to us that the EU regulates Artificial Intelligence providers in a way that is fair to all parts of the ecosystem, including content rightsholders, AI providers, AI application developers and end users.

In particular, we drew the EU AI Office’s attention to the IPTC Photo Metadata Data Mining property, which enables rightsholders to inform web crawlers and AI training systems of the rightsholders’ agreement as to whether or not the content can be used as part of a training data set for building AI models.

The points made are the same as the ones that we made to the IETF/IAB Workshop consultation: that embedded data mining declarations should be part of the ecosystem of opt-outs, because robots.txt, W3C TDM, C2PA and other solutions are not sufficient for all use cases. 

The full consultation text and all public responses will be published by the EU in due course via the consultation home page.

The Internet Architecture Board (IAB), a Committee of the Internet Engineering Task Force (IETF) which decides on standards and protocols that are used to govern the workings of Internet infrastructure, is having a workshop in September on “AI Control”. Discussions will include whether one or more new IETF standards should be defined to govern how AI systems work with Internet content.

As part of the lead-up to this workshop, the IAB and IETF have put out a call for position papers on AI opt-out techniques.

Accordingly, the IPTC Photo Metadata Working Group, in association with partner organisation the PLUS Coalition, submitted a position paper discussing in particular the Data Mining property which was added to the IPTC Photo Metadata Standard last year.

In the paper, the IPTC and PLUS set out their position that data mining opt-out information embedded in the metadata of media files is an essential part of any opt-out solution.

Here is a relevant section of the IPTC submission:

We respectfully suggest that Robots.txt alone is not a viable solution. Robots.txt may allow for communication of rights information applicable to all image assets on a website, or within a web directory, or on specific web pages. However, it is not an efficient method for communicating rights information for individual image files published to a web platform or website; as rights information typically varies from image to image, and as the publication of images to websites is increasingly dynamic.

In addition, the use of robots.txt requires that each user agent must be blocked separately, repeating all exclusions for each AI engine crawler robot. As a result, agents can only be blocked retrospectively — after they have already indexed a site once. This requires that publishers must constantly check their server logs, to search for new user agents crawling their data, and to identify and block bad actors.

In contrast, embedding rights declaration metadata directly into image and video files provides media-specific rights information, protecting images and video resources whether the site/page structure is preserved by crawlers — or the image files are scraped and separated from the original page/site. The owner, distributor, or publisher of an image can embed a coded signal into each image file, allowing downstream systems to read the embedded XMP metadata and to use that information to sort/categorise images and to comply with applicable permissions, prohibitions and constraints.

IPTC, PLUS and XMP metadata standards have been widely adopted and are broadly supported by software developers, as well as in use by major news media, search engines, and publishers for exchanging images in a workflow as part of an “operational best practice.” For example, Google Images currently uses a number of the existing IPTC and PLUS properties to signal ownership, licensor contact info and copyright. For details see https://iptc.org/standards/photo-metadata/quick-guide-to-iptc-photo-metadata-and-google-images/

The paper in PDF format can be downloaded from the IPTC site.

Thanks to David Riecks, Margaret Warren, Michael Steidl from the IPTC Photo Metadata Working Group and to Jeff Sedlik from PLUS for their work on the paper.

The IPTC News Architecture Working Group is happy to announce the release of NewsML-G2 version 2.34.

This version, approved at the IPTC Standards Committee Meeting at the New York Times offices on Wednesday 17th April 2024, contains one small change and one additional feature:

Change Request 218, increase nesting of <related> tags: this allows for <related> items to contain child <related> items, up to three levels of nesting. This can be applied to many NewsML-G2 elements:

• pubHistory/published
• QualRelPropType (used in itemClass, action)
• schemeMeta
• ConceptRelationshipsGroup (used in concept, event, Flex1PropType, Flex1RolePropType, FlexPersonPropType, FlexOrganisationPropType, FlexGeoAreaPropType, FlexPOIPropType, FlexPartyPropType, FlexLocationPropType)

Note that we chose not to allow for recursive nesting because this caused problems with some XML code generators and XML editors.

Change Request 219, add dataMining element to rightsinfo: In accordance with other IPTC standards such as the IPTC Photo Metadata Standard and Video Metadata Hub, we have now added a new element to the <rightsInfo> block to convey a content owner’s wishes in terms of data mining of the content. We recommend the use of the PLUS Vocabulary that is also recommended for the other IPTC standards: https://ns.useplus.org/LDF/ldf-XMPSpecification#DataMining

Here are some examples of its use:

Denying all Generative AI / Machine Learning training using this content:

<rightsInfo>
  <dataMining uri="http://ns.useplus.org/ldf/vocab/DMI-PROHIBITED-AIMLTRAINING"/>
</rightsInfo>

A simple text-based constraint:

<rightsInfo>
  <usageTerms>
    Data mining allowed for academic and research purposes only.
  </usageTerms>
  <dataMining uri="http://ns.useplus.org/ldf/vocab/DMI-PROHIBITED-SEECONSTRAINT" />
</rightsInfo>

A simple text based constraint, expressed using a QCode instead of a URI:

<rightsInfo>
  <usageTerms>
    Reprint rights excluded.
  </usageTerms>
  <dataMining qcode="plusvocab:DMI-PROHIBITED-SEECONSTRAINT" />
</rightsInfo>

A text-based constraint expressed in both English and French:

<rightsInfo>
  <usageTerms xml:lang="en">
    Reprint rights excluded.
  </usageTerms>
  <usageTerms xml:lang="fr">
    droits de réimpression exclus
  </usageTerms>
  <dataMining uri="http://ns.useplus.org/ldf/vocab/DMI-PROHIBITED-SEECONSTRAINT" />
</rightsInfo>

Using the “see embedded rights expression” constraint to express a complex machine-readable rights expression in RightsML:

<rightsInfo>
  <rightsExpressionXML langid="http://www.w3.org/ns/odrl/2/">
    <!-- RightsML goes here... -->
  </rightsExpressionXML>
  <dataMining uri="http://ns.useplus.org/ldf/vocab/DMI-PROHIBITED-SEEEMBEDDEDRIGHTSEXPR"/>>
</rightsInfo>

For more information, contact the IPTC News Architecture Working Group via the public NewsML-G2 mailing list.

Following the IPTC’s recent announcement that Rights holders can exclude images from generative AI with IPTC Photo Metadata Standard 2023.1 , the IPTC Video Metadata Working Group  is very happy to announce that the same capability now exists for video, through IPTC Video Metadata Hub version 1.5.

The “Data Mining” property has been added to this new version of IPTC Video Metadata Hub, which was approved by the IPTC Standards Committee on October 4th, 2023. Because it uses the same XMP identifier as the Photo Metadata Standard property, the existing support in the latest versions of ExifTool will also work for video files.

Therefore, adding metadata to a video file that says it should be excluded from Generative AI indexing is as simple as running this command in a terminal window:

exiftool -XMP-plus:DataMining="Prohibited for Generative AI/ML training" example-video.mp4

(Please note that this will only work in ExifTool version 12.67 and above, i.e. any version of ExifTool released after September 19, 2023)

The possible values of the Data Mining property are listed below:

A corresponding new property “Other Constraints” has also been added to Video Metadata Hub v1.5. This property allows plain-text human-readable constraints to be placed on the video when using the “Prohibited, see plus:OtherConstraints” value of the Data Mining property.

The Video Metadata Hub User Guide and Video Metadata Hub Generator have also been updated to include the new Data Mining property added in version 1.5.

We look forward to seeing video tools (and particularly crawling engines for generative AI training systems) implement the new properties.

Please feel free to discuss the new version of Video Metadata Hub on the public iptc-videometadata discussion group, or contact IPTC via the Contact us form.

Updated in June 2024 to include an image containing the new metadata property

Many image rights owners noticed that their assets were being used as training data for generative AI image creators, and asked the IPTC for a way to express that such use is prohibited. The new version 2023.1 of the IPTC Photo Metadata Standard now provides means to do this: a field named “Data Mining” and a standardised list of values, adopted from the PLUS Coalition. These values can show that data mining is prohibited or allowed either in general, for AI or Machine Learning purposes or for generative AI/ML purposes. The standard was approved by IPTC members on 4th October 2023 and the specifications are now publicly available.

Because these data fields, like all IPTC Photo Metadata, are embedded in the file itself, the information will be retained even after an image is moved from one place to another, for example by syndicating an image or moving an image through a Digital Asset Management system or Content Management System used to publish a website. (Of course, this requires that the embedded metadata is not stripped out by such tools.)

Created in a close collaboration with PLUS Coalition, the publication of the new properties comes after the conclusion of a public draft review period earlier this year. The properties are defined as part of the PLUS schema and incorporated into the IPTC Photo Metadata Standard in the same way that other properties such as Copyright Owner have been specified.

The new properties are now finalised and published. Specifically, the new properties are as follows:

• Data Mining: a field with a value from a controlled value vocabulary. Values come from the PLUS Data Mining vocabulary, reproduced here:
  • http://ns.useplus.org/ldf/vocab/DMI-UNSPECIFIED (Unspecified – no prohibition defined)
  • http://ns.useplus.org/ldf/vocab/DMI-ALLOWED (Allowed)
  • http://ns.useplus.org/ldf/vocab/DMI-PROHIBITED-AIMLTRAINING (Prohibited for AI/ML training)
  • http://ns.useplus.org/ldf/vocab/DMI-PROHIBITED-GENAIMLTRAINING (Prohibited for Generative AI/ML training)
  • http://ns.useplus.org/ldf/vocab/DMI-PROHIBITED-EXCEPTSEARCHENGINEINDEXING (Prohibited except for search engine indexing)
  • http://ns.useplus.org/ldf/vocab/DMI-PROHIBITED (Prohibited)
  • http://ns.useplus.org/ldf/vocab/DMI-PROHIBITED-SEECONSTRAINT (Prohibited, see Other Constraints property)
  • http://ns.useplus.org/ldf/vocab/DMI-PROHIBITED-SEEEMBEDDEDRIGHTSEXPR (Prohibited, see Embedded Encoded Rights Expression property)
  • http://ns.useplus.org/ldf/vocab/DMI-PROHIBITED-SEELINKEDRIGHTSEXPR (Prohibited, see Linked Encoded Rights Expression property)
• Other Constraints: Also defined in the PLUS specification, this text property is to be used when the Data Mining property has the value “http://ns.useplus.org/ldf/vocab/DMI-PROHIBITED-SEECONSTRAINT“. It can specify, in a human-readable form, what other constraints may need to be followed to allow Data Mining, such as “Generative AI training is only allowed for academic purposes” etc.

The IPTC and PLUS Consortium wish to draw users attention to the following notice included in the specification:

Regional laws applying to an asset may prohibit, constrain, or allow data mining for certain purposes (such as search indexing or research), and may overrule the value selected for this property. Similarly, the absence of a prohibition does not indicate that the asset owner grants permission for data mining or any other use of an asset.

The prohibition “Prohibited except for search engine indexing” only permits data mining by search engines available to the public to identify the URL for an asset and its associated data (for the purpose of assisting the public in navigating to the URL for the asset), and prohibits all other uses, such as AI/ML training.

The IPTC encourages all photo metadata software vendors to incorporate the new properties into their tools as soon as possible, to support the needs of the photo industry.

ExifTool, the command-line tool for accessing and manipulating metadata in image files, already supports the new properties. Support was added in the ExifTool version 12.67 release, which is available for download on exiftool.org.

The new version of the specification can be accessed at https://www.iptc.org/std/photometadata/specification/IPTC-PhotoMetadata or from the navigation menu on iptc.org. The IPTC Get Photo Metadata tool and IPTC Photo Metadata Reference images been updated to use the new properties.

The IPTC and PLUS Coalition wish to thank many IPTC and PLUS member organisations and others who took part in the consultation process around these changes. For further information, please contact IPTC using the Contact Us form.