Alibaba bans Claude Code over alleged backdoor risks

-

Alibaba is banning employees from using Anthropic's Claude Code over alleged backdoor risks


Alibaba has told employees to stop using Anthropic's Claude Code because of security concerns. Hidden code in the software could track if a user was in China or linked to a Chinese AI lab, according to Reuters. Starting July 10, employees must use Alibaba's own coding platform, Qoder.

"As Claude Code was recently discovered to carry back-door risks, after comprehensive evaluation, Claude Code has now been added to a list of high-risk software with security vulnerabilities," Alibaba said in an internal notice seen by the South China Morning Post.

The move came after security researchers shared findings on Reddit and GitHub, revealing that a version of Claude Code had been built to examine users' local environments — things like timezone settings and proxy configurations — and quietly embed identifying markers in the data transmitted to Anthropic. Anthropic already prohibits Chinese companies and foreign entities owned by those companies from using its models.

Available Now

Get GoPeek

Preview links without opening tabs. Available on Edge and Firefox. Chrome support coming soon.

Microsoft Edge
Microsoft Edge
Add to Edge — free
→
Mozilla Firefox
Mozilla Firefox
Add to Firefox — free
→
Google Chrome
Google Chrome Soon
Coming to the Chrome Web Store
Demo →

On X, Thariq Shihipar — who works at Anthropic — described the behavior as "an experiment we launched in March that was meant to prevent account abuse from unauthorized resellers and protect against distillation," according to TechCrunch. "The team has landed stronger mitigations since then and we've actually been meaning to take this down for a while," Shihipar said.

The ban arrives amid a broader dispute between the two companies. Anthropic accused Alibaba last month of conducting what it described as the largest known distillation attack on its models to date. In a June 10 letter to two U.S. senators, Anthropic said entities affiliated with Alibaba used roughly 25,000 fraudulent accounts to run nearly 28.8 million exchanges with its models over approximately six weeks. Distillation involves training a weaker model on the outputs of a more powerful one, allowing developers to replicate capabilities without building from scratch.

Anthropic said that distillation effort targeted capabilities including agentic reasoning, software engineering proficiency, and long-horizon task completion. The company urged Washington to pursue stricter semiconductor export restrictions and impose penalties on AI developers caught running distillation operations.

A source cited by Reuters noted that enforcing Anthropic's access restrictions against individuals is difficult, since anyone can spin up a U.S.-based server to disguise where their traffic originates — but organizations like Alibaba are far more sensitive to the legal and compliance consequences of violating such terms. Alibaba has not publicly responded to Anthropic's distillation accusations.

Available Now

Get GoPeek

Preview links without opening tabs. Available on Edge and Firefox. Chrome support coming soon.

Microsoft Edge
Microsoft Edge
Add to Edge — free
→
Mozilla Firefox
Mozilla Firefox
Add to Firefox — free
→
Google Chrome
Google Chrome Soon
Coming to the Chrome Web Store
Demo →

Using GoPeek on GitHub

Comments

Post a Comment

Popular posts from this blog

I bypassed AWS API Gateway auth with a trailing slash. Got $12K bounty.

-
Image
(See original article:  I bypassed AWS API Gateway auth with a trailing slash. Got $12K bounty. ) I was poking at a fintech’s mobile API and noticed something that made no sense. GET /v1/accounts returned 401. GET /v1/accounts/ returned 200 with full account data. One character. Completely different security posture. What I was looking at The API ran on AWS HTTP API — the newer, cheaper alternative to REST API. Lambda authorizer checked a JWT against Cognito, returned an IAM policy. Standard. Routes in OpenAPI: YAML /v1/accounts : get : x-amazon-apigateway-integration : uri : arn:aws:apigateway:... /v1/accounts/{accountId} : get : x-amazon-apigateway-integration : uri : arn:aws:apigateway:... The authorizer ran on every request. But HTTP API makes two decisions: does this route exist, and does the authorizer allow it? Those two layers didn’t agree on what a “match” meant. The weird results I ran ffuf on the path. The results were… inconsistent....
Read more

Former Microsoft dev built a 2.5KB Notepad clone with zero AI features

-
Image
Former Microsoft dev built a 2.5KB Notepad clone with zero AI features Dave Plummer started his programmer's career at Microsoft back in the MS-DOS days. He knows a thing or two about code optimization, which is why he is now trying to strip Notepad of the unwanted AI shovelware Microsoft has been adding over the past few years. Plummer is the man who claims to have created some major Windows features, including the original Task Manager application. Now, the YouTuber and retired developer shared his latest creation: a text editor named TinyRetroPad , which is designed to mimic Notepad's text editing functionality while staying as small as possible on disk. TinyRetroPad is a Notepad-style text editor contained in roughly 2.5 kilobytes (KB). One kilobyte traditionally refers to 1,024 bytes, which means the new editor is definitely using a lot of optimization tricks to contain a usable application inside such a tiny package. Plummer forked ...
Read more

Godot bans AI-generated code contributions

-
Image
Godot says bye bye AI, bans vibe-coded contributions 'We can't trust heavy users of AI to understand their code enough to fix it,' say maintainers who previously called the flood of vibe-coded pull requests 'demoralizing' Vibe coders apparently don't understand what their AI servants write - at least that's what the team behind open-source game engine Godot seems to be implying with a new policy that cracks down on AI-generated contributions. The Godot team announced on Tuesday that they were in the process of rewriting their contribution policy to prohibit almost all use of AI from contributors, citing an overwhelming number of pull requests that have poured in, many of which appear to be AI-generated. Nor, the maintainers suggested, can many heavy AI users be relied on to respond meaningfully to review feedback. "AI cannot take responsibility, and we can't trust heavy users of AI to understand th...
Read more