Technology & Science
Anthropic Alleges 24,000-Account Distillation Theft by DeepSeek, Moonshot AI & MiniMax
On 24 Feb 2026, Anthropic publicly disclosed that three Chinese labs covertly queried its Claude model through thousands of proxy accounts to siphon advanced reasoning and coding capabilities.
Focusing Facts
- Anthropic says ∼16 million Claude interactions were made via about 24,000 fraudulent accounts.
- MiniMax alone generated >13 million queries; Moonshot ~3.4 million and DeepSeek ~150 k, each focusing on Claude’s chain-of-thought and tool-use skills.
- Elon Musk counter-accused Anthropic of prior data theft, citing its 2025 $1.5 billion copyright settlement.
Context
Industrial know-how has always leaked—Samuel Slater’s 1789 emigration with Arkwright’s spinning-frame plans jump-started US textiles, just as 1990s disk-drive code thefts spurred Asian hardware rivals. The alleged mass-API scraping is the first large-scale application-layer replay of that pattern in generative AI, accelerating two long arcs: (i) knowledge diffusion from incumbents to challengers despite IP walls, and (ii) the fusion of intellectual property with geostrategic control after post-2022 chip export bans. If policymakers respond with tighter guardrails and detection regimes, the incident could become the Rosetta Stone case that formalised “model-as-munitions” thinking; if not, it will simply mark another turn in the centuries-old cycle wherein every breakthrough is swiftly copied, setting the stage for the next leap—one that, on a century horizon, may render today’s proprietary chatbots as quaint as water frames and spinning mules.
Perspectives
Business and tech media outlets
Business Standard, Economic Times, CNBC-TV18, Tech Times — Report Anthropic’s claim that DeepSeek, Moonshot AI and MiniMax ran “industrial-scale” distillation campaigns to steal Claude’s advanced capabilities, framing the episode as a serious intellectual-property breach and national-security threat that demands swift government and industry action. Coverage leans heavily on Anthropic’s blog without independent verification, echoing U.S. corporate-security talking points while downplaying Anthropic’s own history of copyright suits and data-scraping practices.
Tech-commentary outlets amplifying Elon Musk and other critics
mint, Digit, The Indian Express — Highlight Musk’s accusation that Anthropic itself ‘stole data at massive scale’ and paid multi-billion-dollar settlements, portraying Anthropic’s outcry over Chinese distillation as hypocritical and ironic. By centering Musk—who is a direct AI competitor—the stories risk echoing a rival’s self-interested spin, shifting focus from the substance of the Chinese firms’ alleged misconduct to sensational claims about Anthropic.
Like what you're reading?