On Monday, February 23, 2026, a major ethical and legal feud erupted in the AI industry. After Anthropic publicly accused three Chinese AI labs (DeepSeek, Moonshot AI, and MiniMax) of stealing Claude’s capabilities through “industrial-scale” attacks, Elon Musk fired back, calling Anthropic out for its own controversial history with data.
Musk’s “hypocrisy” argument centers on the fact that Anthropic—which markets itself as a safety-first, ethical AI company—has faced severe legal consequences for the same behavior it is now decrying in others.
The “Multi-Billion Dollar” Claim
Musk’s assertion that Anthropic has paid billions for “theft” refers to the landmark $1.5 billion settlement reached in late 2025.
- Bartz v. Anthropic (August 2025): Anthropic agreed to pay $1.5 billion to settle a massive class-action lawsuit brought by authors and publishers.
- The “Piracy” Evidence: The court found that Anthropic had used datasets like Books3, LibGen, and PiLiMi to train its Claude models. These are “shadow libraries” containing nearly 500,000 pirated books.+1
- The Ruling: Judge William Alsup ruled that while training on legally acquired data might be “fair use,” downloading and retaining pirated copies was “infringement, plain and simple.”
- Destruction Order: As part of the settlement, Anthropic was legally mandated to destroy all pirated libraries and derivative copies by early 2026.
Anthropic’s Accusation vs. Its Own History
The tech community has noted the irony in Anthropic’s current stance.
| Anthropic’s Complaint (Feb 2026) | The “Hypocrisy” Counter-Argument |
| Accuses Chinese labs of “Distillation Attacks” (using Claude’s outputs to train their models). | Anthropic “distilled” 500,000 pirated books to build Claude’s reasoning capabilities. |
| Claims 24,000 fraudulent accounts were used to scrape its API. | Anthropic was sued by Reddit in 2025 for bypassing anti-scraping systems to harvest 100,000+ posts. |
| Argues “illicit” distillation poses National Security risks. | Musk and others argue that using pirated data is also a violation of the “rule of law” that Western AI labs claim to uphold. |
The National Security Angle
The timing of Anthropic’s blog post—and Musk’s reaction—is particularly sensitive. Anthropic CEO Dario Amodei was reportedly summoned to the Pentagon on the same day the accusations were made.
Anthropic is positioning itself as a strategic U.S. asset, arguing that when Chinese labs “distill” American models, they inherit the intelligence but strip away the safety guardrails (like those preventing the AI from helping build bioweapons). Musk’s retort suggests that a company built on “stolen” data may not be the most reliable arbiter of what constitutes “illicit” training.
“Anthropic is guilty of stealing training data at a massive scale and has had to pay multi-billion-dollar settlements for their theft. This is just a fact.” — Elon Musk, February 23, 2026.
