Platforms Are Splitting on What Counts as Authentic

Substack launched an AI detection tool this week. The Pangram integration scans posts, comments, and replies over 100 characters and shows an estimate of how much was written by a human versus AI. Writers can add an optional AI author's note to disclose their process Substack's new tool tells you who's been writing their newsletters with AI.

Chris Best framed it simply: "This is good use of AI. Software should do everything except the hard part — the idea worth reading, worth sharing."

The same week, Deezer reported that over 50% of daily uploads are now AI-generated experts say exploiting Anthropic's Fable isn't how Kimi K3 got so good. The authenticity crisis isn't coming. It's here, and every platform is handling it differently.

YouTube auto-labels AI videos. TikTok auto-labels. Spotify penalizes AI music. Tidal cuts AI monetization entirely. Patreon blocks training scrapers via Cloudflare. Substack leans into detection with disclosure.

There's no consensus on what "authentic" even means anymore. Some platforms treat AI content as fraud. Others treat it as a signal to label. Substack's bet is that transparency creates trust — and that the hard part of writing (the idea, the caring) can't be automated.

The Anthropic Fable controversy makes this sharper. Kimi K3, Moonshot's new open-weight model, was accused of distilling Fable just two weeks after Fable went public. Experts say that timeline is impossible for full distillation. The training regime shifted to RL. Chinese labs have legitimate technical chops. But the accusation alone — from the White House science advisor, echoed by Treasury — shows how fast the "who owns AI output" fight is escalating.

Patreon's anti-scraping move two weeks ago was the first platform-level line in the sand. Substack's detection tool is the second. They're different approaches to the same problem: when the cost of generating content approaches zero, the value shifts entirely to provenance.

I keep coming back to Best's line: you do want the person to do the hard part. The idea worth reading. That's the only thing AI hasn't commoditized — and it might be the only thing that never will.