Anthropic accuses Alibaba of stealing 29M Claude conversations
Largest known model-theft attack targets Claude's engineering reasoning via 25,000 fake accounts.
This week's AI newsletter leads with Anthropic accusing Alibaba's Qwen lab of orchestrating the largest adversarial distillation attack against Claude. Between April and June, operators used roughly 25,000 fraudulent accounts to run nearly 29 million exchanges, systematically harvesting Claude's software engineering and agentic reasoning capabilities. Anthropic sent evidence to the White House and US senators, marking the first time it publicly named a major Chinese tech giant as the source. Separately, Google's AI talent exodus continued as four senior Gemini contributors (including Jonas Adler, Alexander Pritzel, and DeepMind's John Jumper) announced departures to Anthropic, motivated by pre-IPO equity.
Other major stories include a GitHub supply chain vulnerability where a single anonymous pull request could hijack 300+ high-impact repos like Microsoft's Azure Sentinel and Google's AI Agent Dev Kit. Researchers at Cornell showed that a 13-word planted snippet on Reddit or Wikipedia can reliably steer ChatGPT and Google's AI search into repeating spam. Meanwhile, Europe's AI Act Article 50 goes live August 2, requiring chatbots and deepfakes to disclose their AI nature. The newsletter also highlights that engineering hiring at major tech firms is down only 11% from 2019 (vs 25% for all roles), debunking the AI job apocalypse narrative.
- Anthropic documented 25,000 fake Alibaba accounts extracting 29 million Claude interactions, targeting software engineering and agentic reasoning.
- Google lost four senior Gemini researchers to Anthropic in six days, driven by pre-IPO equity at the startup.
- EU AI Act's Article 50 takes effect August 2, mandating disclosure labels for all generative AI outputs and deepfakes.
Why It Matters
Model theft escalates to unprecedented scale, talent wars reshape major labs, and global regulation tightens.