claude anthropic ai-safety distillation-attacks

Anthropic announces proof of distillation at scale by MiniMax, DeepSeek,Moonshot

twitter.com

February 23, 2026

1 min read

🔥🔥🔥🔥🔥

55/100

Summary

DeepSeek, Moonshot AI, and MiniMax conducted industrial-scale distillation attacks on Anthropic's models, creating over 24,000 fraudulent accounts. These labs generated over 16 million exchanges with Claude to extract its capabilities for training and improving their own models.

Key Takeaways

DeepSeek, Moonshot AI, and MiniMax conducted industrial-scale distillation attacks on Anthropic's models.
These labs created over 24,000 fraudulent accounts to generate more than 16 million exchanges with Claude.
The exchanges were used to extract capabilities from Claude to train and improve their own models.

Read original article

View story →

Coverage & discussions

Hacker News · twitter.com

Anthropic announces proof of distillation at scale by MiniMax, DeepSeek,Moonshot ↗

Community discussions

Hacker News discussion ↗

Historical count: 5 · 5 saved in sample · Historical sample

Analysis ready

Community Sentiment

Mixed

Positives

The release of models by MiniMax, DeepSeek, and Moonshot for free democratizes access to AI technology, fostering competition and innovation in the market.
The ability of Chinese labs to make advanced models accessible for fine-tuning and private hosting enhances opportunities for smaller businesses and research labs.
The feasibility of distilling models with around 16 million sessions suggests that AI-to-AI learning could significantly reduce the resource burden compared to traditional human learning.

Concerns

Anthropic's ongoing legal battles over unauthorized use of proprietary content raises serious ethical concerns about their practices in model training.
The hypocrisy of major AI labs claiming to build AI for everyone while releasing expensive, closed-source models undermines trust in their intentions.
The use of fraudulent accounts to extract capabilities from Claude highlights the questionable methods employed by these labs, potentially eroding the integrity of AI development.