Meta unveiled Muse Image, its latest generative‑AI model, on Tuesday, extending the company’s push to embed artificial intelligence into everyday social‑media experiences. The model, described in a company blog as the first creative offering from the Muse Spark family, can generate new pictures, edit existing ones and combine multiple visual elements into a single composite. Users can access Muse Image today through the Meta AI app, Instagram and WhatsApp, with Meta saying it plans to bring the feature to Facebook, Messenger and its advertising products later this year.
CEO Mark Zuckerberg demonstrated the model’s capabilities during an Instagram Stories post that showed a series of AI‑enhanced effects. The demo featured dozens of Zuckerberg clones, a 360‑degree camera view with AI lead Alexandr Wang and an “exposure portrait” mode starring CTO Andrew Bosworth. The visual showcase highlighted the model’s ability to produce realistic variations on a single subject, as well as more whimsical edits such as turning a user into a pirate.
Beyond novelty, the rollout has sparked a debate over privacy and the potential for deepfake misuse. Meta’s own documentation notes that any user with a public Instagram account can be "tagged" in an AI prompt, allowing the model to generate hyper‑realistic images that incorporate that person’s likeness. In a test conducted by CNET, a public‑account holder named Abrar Al‑Heeti was turned into a pirate in under a minute simply by including her Instagram handle in the prompt. When the same request was made for a private account, the model refused to comply.
To address the risk, Meta introduced a new toggle in Instagram’s Settings under Sharing and reuse. Users can turn off the option labeled "Allow people to reuse your content on Instagram and with AI features at Meta," which blocks anyone from using their public posts in AI‑generated images. The control applies separately to posts and reels, and private accounts are automatically protected because their content is not publicly accessible.
The likeness‑management flow in the Meta AI app also gives users granular control. When a user first attempts to tag themselves in an AI image, the app asks them to confirm their identity with a selfie and optionally three additional photos. After verification, the user can choose who may use their likeness—ranging from "only yourself" to "followers you approve," "mutuals" or "everyone." These settings are intended to give creators, influencers and everyday users a way to safeguard the visual representation of their personal brand.
Meta asserts that Muse Image includes built‑in protections against illegal, abusive or defamatory content. The company points to internal safety layers that filter out disallowed requests. Critics, however, note that previous AI products—such as OpenAI’s ill‑fated Sora video generator—have shown that determined bad actors can sometimes bypass safeguards. The effectiveness of Meta’s new controls will likely be tested as the tool gains traction among the platform’s 2 billion‑plus users.
Industry observers see Muse Image as a strategic move to keep creators within Meta’s ecosystem. By offering AI‑powered editing directly in Instagram Stories and WhatsApp, Meta hopes to reduce the appeal of third‑party creative tools. At the same time, the company must balance innovation with responsibility, especially as the line between user‑generated content and AI‑generated imagery continues to blur.
For now, anyone with a public Instagram profile should review the new privacy toggle and consider whether to limit AI access to their likeness. As the technology evolves, the conversation about consent, deepfakes and platform accountability is only beginning.
Cet article a été rédigé avec l'assistance de l'IA.
News Factory APP - actualités agentiques pour booster votre SEO et AEO.