A viral clip featuring a Brazilian content creator using Kling AI's 2.6 Motion Control has set social media ablaze. The video shows Eder Xavier seamlessly swapping his face and body with those of Stranger Things stars Millie Bobby Brown, David Harbour, and Finn Wolfhard. Viewed over 14 million times on X, the clip caught the eye of a16z partner Justine Moore, who noted, "We’re not prepared for how quickly production pipelines are going to change with AI." She added that these models have "immediate implications for Hollywood," enabling "endless character swaps at a negligible cost."
As tools like Kling, Google’s Veo 3.1, FaceFusion, and OpenAI’s Sora 2 make high-quality synthetic media more accessible, experts warn the technique is primed to spread. Emmanuelle Saliba, Chief Investigative Officer at GetReal Security, stated, "The floodgates are open. It’s never been easier to steal an individual's digital likeness—their voice, their face—and now, bring it to life with a single image. No one is safe." She predicts systemic abuse, from social engineering to disinformation campaigns.
Yu Chen, a professor at Binghamton University, explained that full-body character swapping is a significant escalation over earlier face-only deepfakes. These systems must handle pose estimation, skeletal tracking, clothing transfer, and natural movement synthesis across the entire human form. Chen noted that while financial fraud is a concern, the most immediate harm vector is non-consensual intimate imagery, as the technical barrier for creating synthetic explicit content drops.
Both Saliba and Chen highlighted other threats, including political disinformation and corporate espionage. Scammers could impersonate employees or CEOs, release fabricated "leaked" clips, or bypass security controls by using a believable video persona to gain access to critical systems.
While it's unclear how studios or the actors portrayed will respond, Chen emphasized that developers play a crucial role in implementing safeguards, though responsibility should be shared across platforms, policymakers, and users. He urged researchers to prioritize detection models that identify intrinsic statistical signatures of synthetic content, and for platforms to invest in automated detection and human review. "The rapid democratization of these capabilities means that response frameworks developed today will be tested at scale within months, not years," Chen said.
Meanwhile, other AI developments continue. Alibaba's Tongyi Lab Z-Image Turbo, a 6-billion-parameter model, was released last week, promising state-of-the-art quality on consumer hardware. Developers quickly started cranking out custom fine-tuned adaptations (LoRAs) at a pace outstripping Flux2. The model's efficiency is a key selling point, as competitors like Flux2 demand more resources.
In a separate move, X announced it is restricting Grok's image generation and editing features to paid users after the chatbot was used to create non-consensual sexualized images of real people, including minors. The X Safety account posted that technical limits were added to limit how users can edit images of real people through Grok.
Additionally, Baidu updated its ERNIE model, with ERNIE-5.0-0110 scoring 1,460 points on LMArena's Text leaderboard, landing at #8 globally and becoming the sole Chinese model in the top 10. It ranked ahead of OpenAI's GPT-5.1-High and Google's Gemini-2.5-Pro, and claimed the #2 spot globally for mathematical reasoning.
Amidst this AI proliferation, actor Matthew McConaughey has secured eight trademarks for his iconic "Alright, alright, alright" catchphrase from the 1933 comedy "Dazed and Confused," including a sound mark, as Hollywood grapples with AI's boundaries.
