AI Deepfakes Entice Gay Men with Fake Thirst Traps
An AI-generated account presented as "Derek Lam" amassed more than 31,000 X followers and nearly 40,000 Instagram followers by posting brief, wordless shirtless dancing clips and promoting paid "exclusive" subscription content, according to a June 30, 2026 Vox investigation. The account's X profile photo traced back to a real photo of a different man taken three years earlier, and most followers reportedly did not notice signs of AI generation. Similar accounts documented separately, including @leo.boy2005 (per drag artist Jonnie Reinhart, who says it deceived at least 75,000 followers) and "Alex Silver" (roughly 88,000 followers, monetized via Fanvue), show this is a repeatable pattern. For platform-safety teams, short silent clips leave few forensic cues for detection tooling to catch.
For platform-safety and synthetic-media detection teams, the recurring pattern across these accounts is more instructive than any single case: near-zero audio, sub-10-second clips, and subscription-driven monetization combine to minimize the forensic surface area available for automated or human review, while follower-growth incentives reward scale over disclosure.
What happened
Vox reported June 30, 2026 that an account presented as "Derek Lam" had more than 31,000 followers on X and nearly 40,000 on Instagram, posting brief shirtless dancing clips with no spoken audio and promoting paid "exclusive" subscription content. Vox found the account's X profile photo dated to three years earlier and appeared to show a different man, and reported that most followers did not notice signs the account could be AI-generated. Separately, drag and makeup artist Jonnie Reinhart documented another account, @leo.boy2005 (dubbed "Leo"), which he said had convinced at least 75,000 people it was a real person despite a disclosure buried in its bio; and creator Art Bezrukavenko's "Alex Silver" persona grew to roughly 88,000 Instagram followers and generates income through the AI-content platform Fanvue.
Technical context
Across the accounts reported so far, a common production pattern emerges: short, often silent, vertical video clips that avoid the audio-visual coherence checks (lip-sync, voice consistency) most deepfake detectors rely on. Visual artifacts do surface on close inspection - inconsistent stated birth years, mismatched profile-photo history, physically implausible object motion - but casual scrollers rarely scrutinize a few-second clip closely enough to catch them.
Industry context
Platform AI-content disclosure tools remain inconsistent. Some creators self-disclose in bio text, often behind a "more" link most viewers never click; Instagram's "AI info" label, when applied, is easy to miss and does not appear on every synthetic post. Monetization for these accounts typically flows through subscription platforms built for AI-generated creators, such as Fanvue, rather than through direct platform ad revenue.
For practitioners
Detection signals reported across cases include absent or synthetic audio, physically inconsistent object interactions (props that move independently of the body), profile ages that drift year over year, and reverse-image matches to older, unrelated photos. Moderation and trust-and-safety teams building triage tooling should weigh follower-growth velocity against interactive engagement, since the synthetic accounts documented so far show high passive engagement (likes) relative to substantive interaction.
What to watch
Watch for whether platforms tighten AI-disclosure label placement and consistency, whether subscription platforms serving AI-generated creators face pressure to verify creator identity, and whether detection tooling adapts to the short-clip, no-audio format that current cases show evades casual and some automated review.
Key Points
- 1An AI-generated account posing as Derek Lam drew over 31,000 X and 40,000 Instagram followers with wordless dancing clips and paid subscriptions, per Vox.
- 2Similar accounts documented separately, including one accused of deceiving 75,000 followers, show this is a repeatable pattern rather than an isolated case.
- 3Short, silent clips minimize forensic cues for detectors, so platform-safety teams should weigh engagement quality and reverse-image checks over raw follower counts.
Scoring Rationale
A single Vox investigation (unreachable this audit, blocked domain) documents one AI-generated persona monetizing via paid subscriptions; corroborated as a repeatable pattern by three independently verified accounts of similar gay-targeted AI personas (so.gay, Instinct Magazine). Relevant for synthetic-media detection and platform-safety practitioners, but score reflects single-source status of the headline account's specific figures and a niche audience focus.
Sources
Primary source and supporting public references used for this report.
Practice interview problems based on real data
1,625 SQL & Python problems across 15 industry datasets — the exact type of data you work with.
Try 250 free problems
