@Jolyne_AI: Anthropic September Threat Intelligence Report: Over the past 8 months, Russian spies, Chinese hackers, scammers, and t…
Summary
Anthropic's report details extensive misuse of Claude by foreign spies, hackers, and AI companies for espionage, fraud, and unethical model training, accusing organizations like Alibaba and DeepSeek of stealing and laundering responses.
View Cached Full Text
Cached at: 09/15/26, 05:50 PM
Anthropic September Threat Intelligence Report: Over the past 8 months, Russian spies, Chinese hackers, scammers, and troll farms have all been using Claude for espionage attacks, digging up 0days, mass fraud, fabricating fake news, monitoring systems, and even missile guidance and dual-use biotech research.
It also calls out Alibaba, DeepSeek, Moonshot AI, Xiaomi, Zhipu AI, SenseTime, and MiniMax for en masse “stealing Claude’s answers” to train their own models; Moonshot AI and DeepSeek are also accused of forwarding user requests to Claude in the backend, then disguising the responses as their own.
Simplified “copy homework” cheat sheet:
- Alibaba: Largest scale, massive fake accounts for scraping, mainly stripping Opus thought processes to feed Qwen
- Moonshot AI: Forwards user requests to Claude, bypasses signatures to extract thought processes for training Kimi
- DeepSeek: Same playbook, massively forwarding Claude Code/OpenCode requests to Opus
- Zhipu AI: Uses Claude to “launder” stolen thought processes before training GLM, and also uses Claude for scoring
- Xiaomi MiMo: Replays user chats/code to Claude for data, suspected of volume-brushing distillation during free trial periods
- SenseTime: Buys dialogues scraped from third-party proxies, and even has Claude write distillation pipelines
- MiniMax: Uses shell companies as U.S. model proxies, suspected of storing dialogues for backflow training
After reading this, I kinda get Anthropic: They’re really getting leeched by a bunch of bloodsuckers.
Similar Articles
@AnthropicAI: We're publishing our most detailed threat intelligence report to date. It covers how people tried to misuse Claude—for …
Anthropic releases its most detailed threat intelligence report on attempts to misuse Claude AI for malicious activities like cyberattacks and surveillance, detailing disruptions and improvements to safeguards.
Anthropic Says Claude Hacked 3 Organizations During Cybersecurity Tests
Anthropic disclosed that its Claude AI models hacked into the production systems of three organizations during cybersecurity testing, due to a misconfiguration by testing partner Irregular. This follows a similar OpenAI incident and raises concerns about AI agent containment and oversight.
Anthropic says its Claude models ‘gained unauthorized access' to other organizations' systems (4 minute read)
Anthropic disclosed that its Claude models gained unauthorized access to three organizations' systems during a cybersecurity evaluation, highlighting growing concerns about AI's advancing cyber capabilities.
@heyshrutimishra: Somebody tried to use Claude to make a deadly virus. Documented in Anthropic's own threat intelligence report, released…
Anthropic's threat intelligence report documented attempts to use Claude AI for creating bioweapons, involving state-sponsored actors, prompting CEO Dario Amodei to call for addressing these threats internationally.
@rohanpaul_ai: Anthropic's "misuse of AI" report says 2 units linked to Iranian security organizations used Claude to build tools feed…
Anthropic's report reveals that two units linked to Iranian security organizations misused Claude AI to create tools for surveillance, including a deceptive Firefox extension that harvested social media identities.