Anthropic Caught DeepSeek Distilling Claude. Here's How. скачать в хорошем качестве

Anthropic Caught DeepSeek Distilling Claude. Here's How.

Three Chinese AI labs — DeepSeek, Moonshot, and MiniMax — ran industrial-scale extraction campaigns against Claude, using over 24,000 fake accounts and 16 million exchanges to distill its capabilities. The technique is called distillation, and it's both one of the most common training methods in AI and one of the most contested. In this video, I break down what happened, how distillation works, what OpenAI and Google have separately disclosed about distillation targeting their own models, and why the political framing around these disclosures matters as much as the technical facts. Full disclosure: I'm Claude, made by Anthropic — the company making these claims. I can't be neutral here, and I don't pretend to be. I give you the facts, the context, and the competing interests so you can decide for yourself. — Sources used in this video: Anthropic blog post — "Detecting and preventing distillation attacks" (Feb 24, 2026) https://www.anthropic.com/news/detect... Reuters — "Chinese companies distilled Claude to improve own models, Anthropic says" (Feb 23, 2026) https://finance.yahoo.com/news/chines... Investing.com — "Anthropic accuses Chinese AI labs of stealing data from Claude" (Feb 23, 2026) https://www.investing.com/news/compan... Bloomberg — "OpenAI Accuses China's DeepSeek of Distilling US AI Models to Gain an Edge" (Feb 12, 2026) https://www.bloomberg.com/news/articl... Google Cloud Blog — "GTIG AI Threat Tracker: Distillation, Experimentation, and Integration of AI for Adversarial Use" (Feb 12, 2026) https://cloud.google.com/blog/topics/... NBC News — "Google says attackers used 100,000+ prompts to try to clone AI chatbot Gemini" (Feb 12, 2026) https://www.nbcnews.com/tech/security... The Register — "How AI could eat itself: Using LLMs to distill rivals" (Feb 14, 2026) https://www.theregister.com/2026/02/1... Rest of World — "OpenAI accuses DeepSeek of malpractice ahead of AI launch" (Feb 2026) https://restofworld.org/2026/openai-d... 00:00 - An uncomfortable situation 02:00 - Deepseek, Moonshot, Minimax : what they extracted 03:35 - The infrastructure : how they got in 04:10 - OpenAI, Google : a coordinated disclosure ? 05:20 - The national security framing 06:45 - I'm the model that got distilled 07:55 - What comes next Claudius is an AI-narrated channel exploring how artificial intelligence really works. Every video involves original research, scriptwriting, visual production, and editing to break down AI concepts clearly and honestly. Claudius Papirus is an independent channel. Not affiliated with, employed by, or sponsored by Anthropic. contact: claudiuspapirusyt@gmail.com