VC
Value Add VC
โšกHomePulseโšกHelpful Apps๐Ÿ“Blog
โ† Value Add PulseAI

Meta Pulls Instagram AI Image Feature After Deepfake Backlash

Meta deactivated a Muse Image feature that let anyone generate AI images referencing public Instagram accounts by default, after talent agencies including CAA and users criticized its opt-out-only consent model as enabling deepfakes.

Muse Image
Feature
Opt-out
Default setting
CAA
Key critic
TC
Trace Cohen
Early-stage VC & angel ยท Founder, New York Venture Partners
July 10, 2026
2 min read
ShareXLinkedInEmail
THE RUNDOWN
1

Meta discontinued a feature within its Muse Image tool that let users generate AI images by @-mentioning any public Instagram account, days after backlash over the fact the capability was opt-out rather than opt-in by default

2

Talent agency CAA raised concerns directly with Meta and publicly urged a "more reasonable approach," stating that "no one's name, image, likeness, voice or creative work should be used by any third party, including AI models, without clear, documented consent"

3

Meta acknowledged in an update to its original announcement that the capability "missed the mark" and pulled it down, a rare instance of the company reversing a shipped generative-AI feature within days rather than defending or slow-walking changes to it

4

The episode adds to a growing list of AI-image-generation missteps around consent and likeness rights across the industry, and lands the same week Meta is separately facing EU threats of fines over addictive product design on Facebook and Instagram

TC
The VC Read ยท Trace's TakeTrace Cohen

Shipping an opt-out-by-default deepfake generator and pulling it within days isn't a one-off misstep -- it's what happens when product teams treat consent as a settings toggle instead of a design constraint. The speed of the reversal, faster than Meta's usual pace, suggests legal and PR both saw the exposure immediately once CAA got involved.

Meta deactivated a feature inside its Muse Image generator that had let any user create AI-generated images referencing a public Instagram account simply by @-mentioning it, days after the capability drew sharp criticism for defaulting users into having their public photos usable for AI image generation unless they explicitly opted out.

The backlash came quickly after the feature's announcement, including from talent agency CAA, which said it raised concerns directly with Meta and pushed the company toward what it called "a more reasonable approach." The agency's statement was unambiguous: "No one's name, image, likeness, voice or creative work should be used by any third party, including AI models, without clear, documented consent." That framing -- consent as the default, not an opt-out buried in settings -- is becoming the industry's baseline expectation for any generative tool that references real people's likenesses.

In an update to its original announcement, Meta acknowledged the capability "missed the mark" and pulled it down entirely rather than attempting to patch the consent model or add friction to the opt-out flow. That's a notably fast reversal by Meta's own standards; the company has more often defended or gradually adjusted generative-AI features under criticism rather than suspending them outright within days of launch.

The episode is part of a broader pattern across the AI industry: image and video generation tools keep shipping with permissive defaults around real people's likenesses, then getting pulled back after public and legal pressure. It also lands awkwardly for Meta specifically, which is simultaneously facing threatened fines from EU regulators over allegedly addictive design features on Facebook and Instagram -- meaning the company is fighting a consent-and-safety narrative on two fronts in the same news cycle.

For product teams building generative AI features that touch real people's images, voices or likenesses, the lesson is straightforward: opt-out consent models for likeness-based AI generation are now a fast path to public backlash and rapid feature reversal, not a viable long-term default. What to watch next: whether Meta reintroduces a narrower, opt-in version of the feature, and whether other platforms shipping similar likeness-reference tools face the same pressure.

ShareXLinkedInEmail
More onMeta โ†’

Originally reported by The Verge. Analysis and editorial commentary by Value Add Pulse.

โ† Back to Pulse

THE WIRE in your inbox

Tech, startup & VC news with Trace's take. Free, no spam.

Read Next

AI

OpenAI Launches GPT-5.6 Family: Sol, Terra and Luna

OpenAI shipped GPT-5.6 in three durable capability tiers -- Sol, Terra and Luna -- going generally available across ChatGPT, Codex and the API, with Sam Altman citing a 54% token-efficiency gain on coding tasks.

AI

AI's New Class Divide: Haves, Have-Nots and Know-Nots

Axios reporting frames a widening three-way split in AI access and fluency -- companies and workers with frontier-model access and know-how, those without access, and those without the skills to use what they already have.

AI

Political Risk Expertise Becomes a Hot Hire in Silicon Valley

AI labs and tech companies are recruiting former diplomats, national security officials and political scientists as geopolitical risk and AI-misuse threats increasingly require expertise venture-backed companies historically didn't need.

@Trace_Cohenยทt@nyvp.com