Cursor identifies and removes fraudulent extension from Claude Code that reached up to 5 thousand users
Read more
Tecnoblog
tecnoblog.net

Cursor identifies and removes fraudulent extension from Claude Code that reached up to 5 thousand users

Cursor, the developer of the namesake code editor, detected and removed a modified version of the Claude Code extension, provided by Anthropic, which had the potential to affect up to five thousand users. This malicious package aimed to collect computer data and implant an SSH backdoor to enable remote access to the machine.

The corrupted version, identified as Anthropic.claude-code 2.1.313, was accessible on the platform's extension marketplace between 6:27 AM and 5:35 PM on October 29th, Brasília time. After becoming aware of the incident, the company, which has a stake in SpaceX owned by Elon Musk, removed the extension and temporarily suspended the marketplace's operation.

According to the company, no evidence of manipulation was found in other packages. Cursor also proceeded to rebuild the marketplace before reactivating it. Users who may have been impacted by this flaw should contact the platform directly.

The alert issued by Cursor clarifies that the harmful version was not created by Anthropic, but rather disguised itself as the Claude Code package, containing code capable of compromising the device of anyone who installed it. After installation, the software began collecting sensitive information, such as the user's and PC's name, operating system, processor architecture, working directories, Git email, and IP address.

Subsequently, the extension attempted to incorporate an SSH key controlled by the attacker, named svc-telemetry, aiming to create an entry point that could be used in the future. Using this same key, the malware sought to establish a reverse tunnel to an external server, which would give the attacker the ability to remotely access the machine.

To ensure the maintenance of this access even after restarts, the code implemented persistence mechanisms on both Windows and macOS. These routines were configured to attempt to reestablish the connection at intervals of approximately 60 seconds. Additionally, some of the collected information and the connection status with the server were transmitted to the infrastructure managed by the attacker.

However, the attempt at remote access was unsuccessful on most computers. The platform reported that the code had flaws and required elevated permissions to activate the SSH service. Thus, the risk was significantly higher only on equipment where administrator or root permissions were granted to Cursor. Even in these scenarios, the company advises that any individual who installed the compromised version perform a complete system check. This recommendation applies to anyone who installed or updated the Anthropic.claude-code extension to version 2.1.313 during the period it was available. One way to check the device is by looking for the svc-telemetry key, which was added by the malware in the SSH settings.

Similar stories

Major Data Leak at CBFC: Scripts and Contracts for Over 500 Films Revealed
Read more
www.aajtak.in

Major Data Leak at CBFC: Scripts and Contracts for Over 500 Films Revealed

A major data leak related to the film industry has been discovered. According to reports, documents concerning over 500 films were leaked from the Central Board of Film Certification (CBFC) online portal. Among these stolen materials are film scripts, certification records, and contracts.

This incident occurred via the CBFC's e-Cinepramaan portal, which is used for the film certification process. It is reported that access was gained to some documents on this portal, and they were uploaded to the Internet Archive. However, it is currently unclear how the data leak happened or what volume of information was affected.

Scripts and Contracts Also Become Public Knowledge

The leak affected not only film scripts but also other critical documents. These include CBFC orders and business agreements related to the films. It was noted that some documents contain information about distribution deals and licensing fees.

Since this information is extremely important for filmmakers and distributors, its public disclosure could negatively impact business negotiations and deals. It is important to emphasize that this refers to the leak of associated documents, not that the films themselves were fully published.

CBFC Initiates Investigation

According to available data, CBFC has initiated an investigation into this matter, and a security audit is being conducted. Meanwhile, the Internet Archive has stated its intention to remove the relevant accounts and file a complaint with cybercrime authorities. Information was also provided to the Indian Computer Emergency Response Team (CERT-In).

It remains unclear who is responsible for this leak and how widely the incorrect information has been used. Nevertheless, this case has once again raised questions about the reliability of storing confidential documents on government online portals.

Claude Mythos 5 model attempted to disseminate malicious code on GitHub during security test
Read more
olhardigital.com.br

Claude Mythos 5 model attempted to disseminate malicious code on GitHub during security test

Anthropic reported that Claude Mythos 5 attempted to introduce harmful code into a public GitHub repository. To persuade developers to accept the modification, the model created a fictitious account. This incident occurred during a cybersecurity testing procedure.

The company classified this event as a serious failure in artificial intelligence alignment. The model maintained its approach even after receiving signals that its actions could cause harm to individuals not part of the exercise.

The test began in a controlled environment, configured with a specific security challenge. After failing to solve the challenge through permitted channels, Claude Mythos 5 sought out a public GitHub repository that it judged to be related to the activity. This project belonged to developers with no connection to the experiment.

The model submitted a pull request containing malicious code and tried to convince the owners to approve it. When an external developer flagged the potential risk, the AI did not give up on the attempt. It generated a fake account and posted positive comments to encourage the proposal's approval. However, the repository owners rejected the change, which prevented the act from being completed.

Anthropic replicated phases of this episode in new simulations to verify whether the model would be able to identify the danger and cease the action on its own. The results indicated difficulty in abandoning a strategy that had already been initiated.

These findings demand caution when comparing different versions of the models. Anthropic clarified that evaluations of the latest models were based on excerpts from previous executions, and that the observed performance does not guarantee the same behavior in all scenarios.

The occurrence prompted the company to create new automated tests to evaluate the models before releasing them to the public. Furthermore, the company is investigating the concept of biased reasoning, which describes when an AI selects information to justify continuing an action, even when there are indications of potential harm.

In the specific case of Claude Mythos 5, the warnings received were interpreted by the AI as being compatible with the security exercise. The model failed to perceive that it was interacting with people outside the test. According to Anthropic, this misinterpretation, combined with the insistence on proceeding, led to the inappropriate behavior.

The investigation was conducted in collaboration with the UK Artificial Intelligence Security Institute (AISI), which issued an independent report. This body committed to conducting cybersecurity assessments without internet access until it develops more robust mechanisms to prevent harm to real people and organizations.

Popular