← Back

Wire · regulatory

Anthropic keeps coding a case for open-source AI models with latest spying mishap

Published

11 July 2026

Topic

regulatory

Sectors

Generative AI

Geography

China

Source

Read at cybernews.com

Verified

Fusion42 · 11 July 2026 · Fusion42 review

Anthropic embedded code in Claude Code that covertly collected metadata (timezone, proxy use, Chinese URLs, lab affiliations) from Chinese users and attempted to obfuscate it in the binary before developers discovered and publicised the practice. The incident has revived trust concerns around closed-source AI systems and strengthened the case for open-source alternatives.

This Wire brief sits within Fusion42's coverage of Generative AI.

◆ The Wire takeaway

Anthropic's surveillance code just handed open-source AI projects their most credible sales pitch: closed systems spy, open ones don't have to. If you're building an open-source alternative to Claude or ChatGPT, this week you can point to a Fortune 500 company admitting it collects your users' location and behaviour—and that's your differentiator.

Coverage

1 source · 11 Jul 2026

Related on Wire

Topics

Generative AIanthropic-claudeclosed-source-aiuser-surveillanceopen-source-casetrust-violation