Something unexpected just shook the AI world. Anthropic, known for its Claude models, snuck in a secret tracker aimed at users in China. This hidden code quietly gathered user data, stirring a firestorm over surveillance and trust. But once exposed, Anthropic pulled the plug fast. What happened behind the scenes? And why did a security researcher’s discovery trigger a major backlash?
The Secret Tracker Uncovered
Last week, a web developer going by “Thereallo” found a clever but sneaky trick inside Claude Code. Anthropic was using “prompt steganography”—a way to hide tracking code in plain sight. The code sent back details like the user’s timezone, proxy information, and even links to Chinese AI labs. Most users would never spot it.
Anthropic engineer Thariq Shihipar confirmed the tracker was added in March as an experiment. He explained it “was meant to prevent account abuse from unauthorized resellers and protect against distillation.”
Distillation means others copying or mimicking Anthropic’s models without permission. That’s a real threat. Some unauthorized retailers have sold access to Claude models for as little as $12 per month, while official pro subscriptions cost up to $100 monthly. Anthropic clearly wanted to clamp down on that gray market.
Alibaba Strikes Back
This revelation set off alarm bells at Alibaba. The Chinese giant banned all employees from using Claude Code starting July 10. Alibaba called the software “high-risk” with “security vulnerabilities” after discovering these back-door risks.
Alibaba’s internal notice warned staff that Claude Code posed unacceptable security risks. The ban came just days after the tracker’s exposure and Anthropic’s rollback announcement on July 1.
This isn’t just a corporate scuffle. It’s about national security and trust in AI technology. Alibaba’s move highlights how sensitive China is about surveillance and data privacy, especially involving foreign AI tools.
Anthropic’s Response and the Bigger AI Picture
Anthropic didn’t hesitate to remove the tracker once it was exposed. Shihipar admitted, “Anthropic has actually been meaning to take this down for a while.” The company faced pressure from security researchers and the public.
Meanwhile, Anthropic is pushing forward with new AI models. The US Commerce Department lifted export curbs on Anthropic’s Mythos 5 and Fable 5 models. Since June 26, US organizations can access Mythos 5 again. Fable 5 is now available globally.
US Commerce Secretary Howard Lutnick praised Anthropic’s compliance, saying, “Over the past two weeks, we have worked closely with Anthropic to analyze and approve Fable 5 to ensure alignment across the US Government and strengthen America’s leadership in AI.”
Anthropic’s CEO Dario Amodei described AI regulation with a vivid analogy from The Lord of the Rings. He compared the slow pace of AI rules to the Hobbits trying to rally Treebeard, a slow-moving but powerful tree guardian. His point? AI safety is crucial but moves at a deliberate speed.
Anthropic also launched a “HackerOne” program, inviting security researchers to submit jailbreak vulnerabilities. They have a 24/7 team monitoring threats and work with Amazon, Microsoft, Google, and others to create AI safety frameworks.
What’s Next for AI, Trust, and Transparency?
This episode shows how delicate the AI trust ecosystem is. Hidden trackers, even with good intentions like stopping abuse, risk shaking user confidence and sparking international tensions.
Anthropic’s swift rollback and transparency efforts show the AI industry is still learning how to balance innovation and privacy. Alibaba’s ban sends a strong message about data security and sovereignty in China.
Will other AI companies face similar scrutiny? Probably. As AI models spread worldwide, companies must tread carefully. The future demands clear rules, open communication, and respect for user privacy.
One thing’s clear: the AI landscape is changing fast. Tracking codes hidden in plain sight won’t fly for long. Trust is the new currency, and everyone—from innovators to regulators—has skin in the game.
Based on
- Secret Claude tracker shocks users after Anthropic’s anti-surveillance stance — arstechnica.com
- Anthropic rolls back China tracking code | Semafor — semafor.com
- Alibaba bans staff from using Claude Code over Anthropic spyware concerns | South China Morning Post — scmp.com
- After spooking Trump into safety testing, Anthropic AI models get global release – Ars Technica — arstechnica.com
- Alibaba reportedly bans employees from using Claude Code | TechCrunch — techcrunch.com
