Anthropic AI has recently stated that it has detected significant distillation attacks on its AI reasoning models from the Chinese company DeepSeek, along with two other firms, Moonshot AI and MiniMax. In a post shared on X, the company outlined that these entities created over 24,000 counterfeit accounts and conducted more than 16 million exchanges with their AI model, Claude. These actions are said to have allowed them to extract capabilities from Claude to enhance their own models.

The practice of distillation can be used legitimately; for instance, AI laboratories often employ it to generate smaller and more cost-effective AI models for their users. However, Anthropic AI has raised concerns that foreign entities illicitly distilling American models could eliminate necessary safeguards. This could potentially allow these capabilities to be integrated into military, intelligence, and surveillance systems, raising the stakes in the evolving landscape of AI technology.

Increase in Attack Sophistication

Anthropic emphasised that the frequency and complexity of these attacks have been escalating. The company underscored the necessity for urgent and coordinated measures to address the issue, highlighting the roles of industry stakeholders, policymakers, and the wider AI research community in formulating a response.

Background of Anthropic AI

Founded in 2021 by former OpenAI researchers, including siblings Dario and Daniela Amodei, Anthropic AI is focused on developing reliable and interpretable AI systems. The company’s work prioritizes safety, and it is best known for the Claude family of large language models, which are engineered to be useful, transparent, and non-harmful. Their research is guided by a concept known as “constitutional AI,” which utilizes explicit principles rather than solely depending on human feedback.

Overview of DeepSeek

DeepSeek, a Chinese artificial intelligence firm established in 2023, is recognized for creating advanced open-weight large language models and its own chatbot. Models like DeepSeek-R1 and DeepSeek-V3 are noted for their efficient architecture that provides high performance while reducing computational costs when compared to numerous Western counterparts. The DeepSeek chatbot gained rapid popularity in early 2025, being downloaded extensively, due to its robust reasoning abilities, long-context comprehension, and multilingual performance.

Regulatory Concerns and Response

Despite the accolades, DeepSeek’s operations have drawn the attention of regulators and raised national security alarms in various nations, primarily because of issues related to data privacy and the methods employed in training its models. As of now, there has been no response from DeepSeek AI, Moonshot AI, or MiniMax to address the allegations made by Anthropic. The claims bring heightened scrutiny towards DeepSeek’s practices, especially within the context of current international relations.

Presently, there are no regulations in either the United States or China that govern these types of activities, leaving a gap in oversight as the landscape of artificial intelligence continues to evolve rapidly. The growing concerns around data security and ethical implications underscore the need for a regulatory framework to address these emerging challenges within the AI sector.

Long or Short, get news the way you like. No ads. No redirections. Download Newspin and Stay Alert, The CSR Journal Mobile app, for fast, crisp, clean updates!

App Store –  https://apps.apple.com/in/app/newspin/id6746449540 

Google Play Store – https://play.google.com/store/apps/details?id=com.inventifweb.newspin&pcampaignid=web_share