Anthropic has accused three Chinese artificial intelligence companies of conducting “industrial-scale campaigns” designed to “illicitly extract” its technology through what are known as distillation attacks. The company claims these firms, including DeepSeek, the developer of the widely used DeepSeek AI models, created approximately 24,000 fraudulent accounts to obscure their activities. This allegation raises significant concerns regarding intellectual property and national security.
In a detailed blog post, Anthropic stated, “We have identified industrial-scale campaigns by three AI laboratories—DeepSeek, Moonshot, and MiniMax—to illicitly extract Claude’s capabilities to improve their own models.” The company claims these labs engaged in over 16 million exchanges with its Claude model, violating both its terms of service and regional access restrictions.
This isn’t the first time DeepSeek has faced such accusations. Earlier this year, OpenAI also alleged that the company was involved in distillation attacks, essentially stealing proprietary technology. While many observers mocked OpenAI for its claims, citing the industry’s reliance on copyrighted materials, the ongoing disputes highlight a troubling paradox. AI companies, including OpenAI, have defended their practices of training models on copyrighted works, arguing that they have no choice given the competitive landscape, particularly regarding Chinese firms’ disregard for intellectual property rights.
During an AI event in July 2025, former President Donald Trump stated, “You can’t be expected to have a successful AI program when every single article, book, or anything else that you’ve read or studied, you’re supposed to pay for.” He further criticized China, asserting that the country does not adhere to copyright norms.
The notion of distillation, while a legitimate training technique, can also be exploited for less than ethical purposes. Researchers utilize distillation by running various prompts to understand a model’s responses, which can lead to reverse-engineering aspects of the technology. Anthropic explained, “Distillation is a widely used and legitimate training method. However, it can also be used for illicit purposes: competitors can acquire powerful capabilities from other labs in a fraction of the time and cost compared to developing them independently.” This dual-use nature complicates the discourse surrounding intellectual property in the AI sector.
Chinese companies are frequently criticized for flouting international intellectual property treaties and reverse-engineering technology from Western firms. While Anthropic contends that the distillation attacks it uncovered breached its terms of service, it remains unclear whether any formal international laws were violated or what recourse exists beyond suspending the offending accounts.
In light of these threats, Anthropic is advocating for a coordinated response among AI companies, government agencies, and other stakeholders. The firm warns that the intensity and sophistication of these campaigns are increasing, emphasizing that “the window to act is narrow, and the threat extends beyond any single company or region.” A collaborative effort among industry players and policymakers may be essential to safeguard intellectual property rights.
The AI sector is currently experiencing a significant surge in investment, with billions of dollars dedicated to infrastructure, data centers, and research. Should rival foreign companies successfully replicate advanced large language model technology through distillation, U.S. firms could find themselves at a distinct disadvantage in this rapidly evolving market.
As the landscape of AI continues to evolve, the balance between innovation and the protection of intellectual property rights becomes increasingly critical. The ongoing tensions may shape future regulatory frameworks and influence how AI companies navigate their competitive strategies.
In conclusion, as the industry grapples with these complex challenges, the role of international cooperation and adherence to intellectual property norms will likely become paramount in ensuring fair competition in the global AI marketplace.
See also
Intuit Achieves $4.15 EPS, 25% Growth, as AI-Driven Strategy Hits Inflection Point
Perplexity Launches Perplexity Computer, A Unified AI System for Full Workflow Automation
US Military Uses Anthropic’s Claude in Iran Strikes Despite Trump’s Ban
Microsoft 365 Conference Reveals Key Strategies for Scaling AI with Copilot Sessions
