Meta pulls Instagram AI image tool, testing limits of AI likeness scraping

Just three days after launching, Meta has removed an Instagram feature that allowed users to use AI to create images based on profile pictures of other users. The company’s decision to remove the function shows growing concern about consent and the use of images in the age of artificial intelligence.

The backtracking is an early setback for Meta’s efforts to catch up with OpenAI and Google in the field of generative AI. While the Muse image model is still available, the company has scrapped the option to create images based on public Instagram profiles in its image generation requests after criticism from creators, performers, and privacy advocates.

What Meta built, and what it walked back

Meta unveiled Muse Image on July 7, stating that the new AI-powered image generation tool is the first product of its Superintelligence Lab led by Alexandr Wang. The new technology drives 30 effects on Instagram and is said to surpass Google Nano Banana 2 according to Meta’s internal testing, while trailing only ChatGPT’s image generator.

The dispute focused not on the model as such but on the way it treated the images uploaded by users. By tagging their adult public Instagram account, users could use the public images from this account in AI-generated photos. Users below 18 years of age and the private accounts were excluded from the system, while public adult accounts were automatically enrolled unless disabled manually by the users.

According to Meta, the aim of the feature was to give users “control over whether their public content could be referenced.” However, many critics think just the opposite: introducing the opt-out system means making the assumption that consent is given by default.

As a result, by Friday, the company changed its mind.

A Meta representative stated to Variety that “we’ve heard the feedback that this feature missed the mark, so it’s no longer available.”

Creators and safety groups forced the backtrack

The feature drew criticism almost immediately after its launch. Screen Actors Guild–American Federation of Television and Radio Artists (SAG-AFTRA) advised its members to switch off the feature, warning artists to “take action to protect your likeness.”

The talent agency Creative Artists Agency, LLC (CAA), which represents Tom Hanks and Meryl Streep among others, said it was in direct contact with Meta and expressed that the AI system should never use the names, images, likenesses, and voices of people in the absence of their explicit consent.

Meanwhile, privacy advocates warned that the technology might make impersonation, non-consensual intimate imagery, and online scams easier. Haley McNamara, of the National Center on Sexual Exploitation, said in her interview for The Verge that the notion of individuals opting out transformed the accountability from the company to users of technology.

The dispute highlights the legal battle over AI likenesses

The controversy highlights a broader question facing the AI industry: Does posting a photo publicly also give companies the right to use it for generative AI?

This topic connects more and more with European privacy law. The General Data Protection Regulation (GDPR) defines images of people as personal data, which means organizations need to find a legal basis for processing such data under Article 6.

If there is a need to obtain the consent of an individual, Article 7 says that consent should be freely given, specific, and informed. Article 9 specifies that the processing of biometric data should require more liability in order to uniquely identify a person. Articles 12-14 also state that the organizations should describe how personal information is used so that individuals are more aware of its use.

Regulators are also expanding their focus beyond traditional privacy rules. The EU AI Act requires providers to inform the public if content has been made or augmented using AI technologies, such as deepfakes. The purpose of this requirement is to provide transparency about synthetic media instead of prohibiting the use of this type of technology in its entirety.

But Meta is not the only company facing this issue. Just a few months back, OpenAI was forced to remove a similar opt-out option in its Sora video model because of the backlash it received regarding its use of publicly available images.

The interventions confirm that AI technology developers are facing tougher challenges because of the attention that regulators and creators devote to making sure that consent is given unambiguously before manipulating anyone’s likeness.

The discussion about the issue has already moved into the realm of legislation. California has passed two bills, AB 2602 and AB 1836, requiring permission before the commercial use of an AI-generated image of a living or dead performer.

While the laws may be more about entertainment agreements than about the applicability of AI in the consumer realm, they nonetheless indicate a tendency to consider AI likenesses as things that need permission, rather than just presuming that they can be used by anyone.

Investors seem unaffected by all the fuss. As per previous reports by Cryptopolitan, Meta shares closed at $615.58 on the day of the launch (July 7), which is 2.55% up on the day and is their highest closing in a month.

Still, there remains the question of whether publicity of content allows users to take advantage of it for AI developments. In fact, if Meta’s recent decisions are a reflection of wider trend in the AI market, simply making something public does not entitle others to use it in AI at all. At the same time, it is becoming clear that consent will play an important role in the development of new AI products for consumers as it is increasingly requested by the industry as well as public organizations.