Meta removes AI deepfake profiles posing as celebrities

Meta has removed a network of social media accounts creating synthetic images and videos of high-profile figures, including actress Mia Goth and actor Scarlett Johansson. The takedowns occurred after fake media racked up significant engagement on Instagram. A verified account using the handle @miagoth_ accumulated over 1 million followers before circulating an AI-generated photo of Goth wearing a shirt targeting her ex-husband, which gathered nearly 500,000 likes and populated Google search summaries.

Synthetic media featuring celebrities continues to generate massive reach across Meta platforms. Video clips depicting AI versions of Shrek and soccer star Lionel Messi have drawn 3.1 million and 2.1 million views respectively, while clips featuring former Apple CEO Steve Jobs reached 371,000 views and drivers Lewis Hamilton and Max Verstappen garnered 2.9 million views. Another video of Taylor Swift at the 2009 VMAs accumulated 1.8 million views. Despite these figures, Meta exempts still images from its mandatory AI labeling rules, though it requires labels for realistic synthetic audio and video, alongside profile badges for fully artificial personas. Beyond viral posts, accounts have commercialized deepfakes, selling $29-per-month courses on generating AI content or directing users to paid subscription channels.

Platform moderation practices remain fragmented compared to emerging legal standards. YouTube relies on a proprietary likeness detection tool created alongside the entertainment industry that scans visual and audio matches for all adult account holders, removing over 95 percent of reported videos under the program during the first half of 2026. Nationally, the US Take It Down Act passed in 2025 mandates platforms clear nonconsensual intimate synthetic images within 48 hours. For digital strategists and trust officers, these enforcement gaps show that existing platform guardrails struggle against persistent deepfake distribution, making brand protection and identity monitoring increasingly complex across social networks.