Adobe’s ‘Content analysis’ program could be using your photos and video to train AI without your knowledge

If you’re an Adobe Creative Cloud subscriber, you might want to go and turn off a new setting immediately. It’s been discovered that Adobe has automatically opted users into a ‘Content analysis’ program that allows Adobe to analyze your media files in Creative Cloud for use in its machine learning training programs.

It’s unclear when exactly this program and setting went into effect, but it became widely noticed after Twitter user Krita_Painting posted the following tweet with a screenshot showing off the setting, which was enabled by default in her Adobe Creative Cloud account.

This setting, which appears to be on by default for all Creative Cloud subscribers, allows Adobe to perform content analysis on ‘Creative Cloud and Document Cloud content includ[ing] but [not] limited to image, audio, video, text or document files, and associated data.’ Adobe says in its Content analysis FAQ that files stored locally on your device are not analysed or processed.

As for how Adobe analyzes content, the company says it ‘first aggregate[s] [users’] content with other content and then use[s] the aggregated content to train our algorithms and thus improve our products and services.’ In other words, Adobe is taking users’ media, putting it into a training set used to teach its machine learning features, then bulk processing it.

Adobe says ‘insights obtained through content analysis will not be used to re-create your content or lead to identifying any personal information’ and notes no ‘content is included in our products or services unless you make them public (for example, contributions to Adobe Stock and Behance).’ But that’s not exactly the case, as there are a handful of exceptions that allows Adobe to access content from users’ Creative Cloud libraries, such as participating in Adobe’s Photoshop Improvement Program, Adobe Stock and ‘Certain beta, prerelease, or early access products or features.’

A screenshot of the ‘Content analysis’ feature hidden within the ‘Privacy and personal data’ section of my Adobe Account. This is how the page looked when I first visited, with the setting turned on by default.

There’s also the possibility that users’ content is manually reviewed by people ‘for product improvement and development purposes’ in select circumstances. If users’ content is made public through Adobe Stock, Behance, live-streamed with Adobe Live, submitted to be a feature on Adobe Express or submitted as a tutorial in Lightroom, Adobe gives itself permission to manually review users’ content, with seemingly little restrictions on what can be manually reviewed by humans. Users’ content can also be manually reviewed if they’re using a pre-release, beta or early access program.

This means humans within Adobe (or contracted personnel) could possibly review sensitive media from users’ Creative Cloud files, should a user fall within one of the categories of exceptions mentioned above and on Adobe’s Content Analysis FAQ page. This clearly causes concern for photographers and other creatives whose media involves more sensitive imagery.

Considering that the content analysis program appears to be something Creative Cloud users were automatically enrolled in, creatives from around the world could be unwittingly subjecting their clients and subjects to having another algorithm or human look at images that both parties thought were protected and shielded from anyone other than the consenting parties. It’s clear how this could turn into a privacy nightmare, should someone at Adobe misuse the images they come across or otherwise make private images public.

Even with those explainers from the FAQ page, Adobe never explicitly states what exactly it’s doing with the media taken from its users

Even with those explainers from the FAQ page, Adobe never explicitly states what exactly it’s doing with the media taken from its users, instead skirting the topic by explaining what machine learning is and how exactly the company uses machine learning in its AI-powered tools. Below is the relevant section in the company’s Content analysis FAQ webpage:

‘Adobe primarily uses machine learning in Creative Cloud and Document Cloud to analyze your content. Machine learning describes a subset of artificial intelligence in which a computing system uses algorithms to analyze and learn from data without human intervention to draw inferences from patterns and make predictions. The system may continue to learn and improve over time as it receives more data.’ The FAQ page then goes on to push the benefits AI-powered tools afford Creative Cloud users and details various examples of how Adobe’s AI features in Lightroom, Photoshop and more.

The privacy concerns surrounding this program should be evident. All it takes is a rogue employee or contractor to misuse private images that Creative Cloud users keep in what is believed to be a private Creative Cloud account to run into all kinds of issues.

For now, you can opt out by logging into your Adobe account, tapping on your profile photo in the top-right-hand corner of Adobe’s website, clicking ‘View Account,’ tapping ‘Edit Profile’ under your profile picture, going to ‘Privacy and personal data’ in the left-hand sidebar, then scrolling down to ‘Content analysis’ to turn off the automatic feature. Below is a video walkthrough of the process:

We’ve inquired with an Adobe representative for additional information to clarify some details and will update this article when we receive an informative response.