As 2025 wrapped up, Instagram’s head Adam Mosseri issued a warning about AI, lamenting that authenticity was becoming endlessly replicable. He worried that the very qualities that made creators meaningful—realness, connection, a unique voice—were now accessible to anyone with the right tools. Yet he insisted people still crave genuine content. His answer? A system that labels real media. He proposed that cameras cryptographically sign images at capture, creating a verifiable chain of custody. That system already exists: C2PA, the Coalition for Content Provenance and Authenticity. But despite its existence, Instagram isn’t using it effectively. Instead, the platform is pushing forward with aggressive AI content generation, turning the promise of authenticity into a hollow gesture. AI has become so good at mimicking reality that it threatens the foundation of social media—creators, trends, and trust. It can replicate dance moves, recreate photo shoots, fabricate influencers, and churn out vast amounts of generic, low-quality content. Creators are fighting back by embracing raw, imperfect aesthetics, but AI can now mimic those too. Worse, it’s being used to spread misinformation—like false narratives around the ICE protests in Minnesota or the tragic deaths of Renee Nicole Good and Alex Pretti. Over the past few years, major tech companies have publicly backed C2PA as a solution. Founded in 2021 by Adobe, Intel, Microsoft, ARM, Truepic, and the BBC, C2PA embeds invisible metadata into media at creation or editing, tracking origin, edits, and AI involvement. Meta joined the C2PA Steering Committee in September 2024, calling content provenance “critical” for digital health. The standard is also supported by Google, OpenAI, TikTok, Qualcomm, and others. But adoption remains weak. Camera makers like Canon, Nikon, Sony, FujiFilm, and Leica have rolled out C2PA support slowly and inconsistently, mostly on new models. Older cameras still produce valid photos, but without provenance, trust relies on context and reputation—something that breaks down online. Worse, C2PA metadata can be stripped away accidentally or on purpose. Platforms like LinkedIn and TikTok fail to reliably display C2PA labels. YouTube uses C2PA, Google’s SynthID, and other tools, but the labels are inconsistent, hard to find, and often absent on desktop. On Instagram, AI labels—now called “AI info”—are buried in tiny text below accounts, sometimes replaced by song names, and not always visible at all. To see a label, users must manually open a three-dot menu or use a browser extension or C2PAchecker website. Most people don’t know these tools exist. Even when labels appear, they’re not always accurate. Meta’s earlier attempt to label AI content backfired, angering photographers who saw real photos incorrectly flagged. Now, the labels are nearly invisible. C2PA is not a silver bullet. As Adobe’s Andy Parsons admitted, it solves a class of problems, not all. And it depends on universal participation—something no one can enforce. X, formerly Twitter, was a founding member but left after Musk took over. The platform now actively hosts AI-generated content, including violent and sexualized deepfakes, with no meaningful safeguards. That’s a massive blind spot in any authenticity effort. Reality Defender CEO Ben Colman argues that relying on labeling assumes only certain tools create bad content—a flawed assumption. Malicious AI isn’t limited to a few platforms. And research shows transparency warnings often fail to stop harm. There’s little evidence that labeling alone protects users. Still, companies keep calling C2PA a “step forward.” But progress is slow, and the incentives are misaligned. OpenAI profits from subscriptions that unlock higher AI generation limits. YouTube’s fastest-growing channels in July 2024 were dominated by AI slop, despite policies against inauthentic content. Meta is planning to monetize AI features on Instagram, Facebook, and WhatsApp. Mark Zuckerberg continues to frame AI as the inevitable future of social media. Platforms aren’t just failing to stop AI slop—they’re profiting from it. Like controversial or inflammatory content, AI-generated material drives engagement, keeps users on the platform longer, and fuels ad revenue. It’s not just harmful—it’s profitable. Some platforms are shifting focus from content to creators. Instagram may soon prioritize who says something over what is said. YouTube, after the deaths of Good and Pretti, began promoting official news sources and previews during breaking events. But Google is simultaneously replacing real headlines with AI-generated summaries—often inaccurate. The truth is, any system that prevents AI from being mistaken for real content contradicts the business models of the companies that built it. Why would Meta, Google, or OpenAI limit the tools that generate their revenue? Mosseri’s vision of a world where creators must be “real, transparent, and consistent” assumes people can easily navigate a sea of AI fakery. But if that were simple, CAPTCHAs and community notes would have solved it long ago. The reality is, the platforms that claim to care about authenticity are the same ones flooding the internet with AI slop. The war on reality isn’t being lost—it’s being actively fought for.