Why Big Tech Is Drowning in the AI Slop It Creates

In the final days of 2025, Instagram head Adam Mosseri published a post that is hard to describe as anything but a manifesto of digital anxiety. "Authenticity is becoming infinitely reproducible," he wrote. Everything that made content creators matter — the ability to be real, to connect with an audience, to have a voice that cannot be faked — is now available to anyone with the right tools.

Mosseri proposed an elegant fix: camera makers would cryptographically sign photos at the moment of capture, creating a chain of trust. The result would be a reliable way to determine what is not a product of artificial intelligence. It sounds like a plan.

The problem is that it has sounded like a plan for several years now.

The Verge noted a fundamental contradiction running through the entire industry. The companies that built generative AI tools and keep scaling them are at the same time cast as fighters against the consequences of their own technology. Meta, which owns Instagram and Facebook, is actively developing its own generative models, building AI features into every product and encouraging the creation of synthetic content. And the same executive at the same company publicly mourns the death of authenticity. This is not hypocrisy in the classic sense — it is a structural conflict of interest built into the business model.

Progress on labeling deepfakes and AI-generated content really is dismally slow. The C2PA standard that Mosseri refers to embeds provenance metadata about an image directly in the file. Several camera makers — Sony, Leica, Nikon — are already experimenting with the technology. But a chasm separates a lab prototype from mass adoption. The vast majority of photos in the world are taken on smartphones, and neither Apple nor Samsung has so far integrated C2PA into its devices at a level that could be called an industry standard. What's more, metadata is easily stripped on upload to most platforms — including, ironically, Instagram itself.

There is also a deeper problem. Even if labeling of "real" photos works perfectly, it does not solve the question of content that already exists. Billions of images uploaded before the standard was adopted will remain unsigned. The absence of a cryptographic mark will not mean a photo is fake — it will only mean it was taken before the labeling era or on a device that does not support the standard. A system that cannot tell an old genuine photo from a new fake has limited value.

Meanwhile, the scale of the AI slop problem — low-quality synthetic content filling feeds and search results — keeps growing exponentially. Facebook has already become a place where AI-generated images collect millions of likes and reposts. Google search results are clogged with sites full of AI-written text. Amazon is struggling with a flood of books written by language models. Each of these companies is both victim and source of the problem: they supply the tools to generate content, the platforms to distribute it and the algorithms that reward its engaging qualities.

Critics rightly point out that tech giants have no economic incentive to truly solve this problem. Synthetic content generates engagement, engagement generates ad revenue, and ad revenue is what these platforms exist for. Labeling AI content, if implemented honestly and universally, could reduce interaction with it and thereby hit the metrics shown to investors in quarterly reports. In this light, Mosseri's statements look less like an action plan and more like a PR exercise: name the problem, offer a distant technological solution and keep profiting from the status quo.

The situation resembles tobacco companies funding research on the harms of smoking for decades while continuing to sell cigarettes. The difference is that AI slop does not literally kill — it destroys the information environment, undermines trust in visual evidence and creates a world where telling truth from fiction becomes a task requiring technical expertise. If the biggest tech companies really want to fight this phenomenon, they should start not with cryptographic signatures on cameras but with their own products — by restricting generative tools, labeling transparently on their own platforms, and building algorithms that do not push synthetic content into the feeds of billions of users.

Until that happens, all talk of fighting AI slop will remain exactly what it is — slop of a different kind.

*Meta has been designated an extremist organization and is banned in Russia.