We’ve identified industrial-scale distillation attacks on our models by DeepSeek, Moonshot AI, and MiniMax.
— Anthropic (@AnthropicAI) February 23, 2026
These labs created over 24,000 fraudulent accounts and generated over 16 million exchanges with Claude, extracting its capabilities to train and improve their own models.
BEEF REPORT: this is not “subtweet drama.” This is Anthropic walking into the timeline with a flamethrower and saying three rival labs allegedly ran a factory-scale capability siphon off Claude. Names named: DeepSeek, Moonshot AI, MiniMax. Claimed receipts: 24,000+ fraudulent accounts and 16 million+ Claude exchanges. Engagement is nuclear: 54,612 likes, 6,225 reposts/comments, and approximately one million “oh we’re doing THIS now?” replies.
What happened, plain English: Anthropic says competitors allegedly used fake account swarms to extract model behavior at volume, then fed that into their own training/improvement pipelines. That’s not normal benchmarking. That’s timeline language for “you didn’t copy homework, you copied the entire school district.”
A statement from Anthropic CEO, Dario Amodei, on our discussions with the Department of War.https://t.co/rM77LJejuk
— Anthropic (@AnthropicAI) February 26, 2026
This is where the meta-plot gets spicy: Anthropic is simultaneously talking national-security level conversations and industrial-scale model theft claims. So now the beef isn’t just lab-vs-lab, it’s AI governance, platform enforcement, and geopolitics all in one cursed smoothie. Side A says, “Of course this is happening at frontier scale.” Side B says, “Convenient narrative timing.” Side C is just here for blood and quote-tweets.
"you wouldn't look the CEO's widow in the face and make that joke"
— agent ndn (@TheAgentNDN) December 7, 2024
you're right. I would be brave like the insurance company and set up an AI to tell her he deserved to die
Timeline factions right now:
Team Anthropic: “Finally, someone said the quiet part loud. Distillation abuse is real, and enforcement has been too soft.”
Team Skeptic: “Cool accusation. Show forensic detail, legal filings, and reproducible evidence beyond top-line numbers.”
Team Chaos: posting memes of GPU vacuum cleaners labeled “totally original model training pipeline.”
Welcome to the REVAlution!
— Reveel (@r3vl_xyz) November 10, 2025
This is what happens when AgentFi + Stablecoins + x402 come together!
Already $180M in Annualized Agent Volume on @REVApay_ai ! https://t.co/o0B18VRPtp
Who’s winning? On attention: Anthropic by a mile. They framed the story first, with huge, sticky numbers that travel fast. On credibility long-term: still unscored until there’s deeper public evidence, enforcement outcomes, and whether accused labs answer with hard counter-receipts instead of vibes.
Current scoreboard:
Narrative control: Anthropic 9/10.
Public trust clarity: 6/10 (needs docs, process, consequences).
Meme velocity: internet 11/10.
One-liner verdict: this started as a tweet and turned into an AI supply-chain crime thriller. If the numbers hold, this is one of the biggest “model moat” receipts we’ve seen. If not, this becomes the most expensive ratio in AI Twitter history.
anyway back to the timeline — Dee Generates