Anthropic hacks Claude developer laptops, delivering remote‑access trojan
anthropic claude startup
| Source: HN | Original article
Anthropic's Claude development laptops were compromised after a hacker delivered a remote‑access trojan via a malicious code extension.
Anthropic’s official Claude Code extension was compromised in a supply‑chain attack that delivered a persistent remote‑access trojan to developers’ laptops. Security researchers at OpenSourceMalware identified a tampered package on the Open VSX marketplace – anthropic.claude‑code version 2.1.313 – that installs an SSH key, opens a reverse tunnel and adds startup scripts, giving attackers continuous control of the host. The malicious code targets macOS on Apple Silicon, and the extension is used by tens of millions of developers worldwide to access Claude through Visual Studio Code and Cursor.
The breach matters because it demonstrates how front‑line AI tooling can become a vector for large‑scale compromise. By inserting the backdoor into a trusted extension, threat actors bypassed traditional perimeter defenses and gained footholds on development machines that often hold proprietary code and credentials. The incident also raises questions about the security of the broader AI ecosystem, where rapid deployment of third‑party plugins and extensions is commonplace.
Anthropic’s Threat Intelligence team has acknowledged the case and said it is working with its Safeguards group to improve defenses. The company has not yet disclosed whether any data was exfiltrated or how the attackers gained initial access to the repository. Users are being urged to uninstall the affected version and verify the integrity of their development environments.
Going forward, watch for Anthropic’s official remediation steps, including a patched release of Claude Code and any guidance on supply‑chain hardening. Security analysts will likely scrutinise other AI‑related extensions for similar tampering, and regulators may intensify oversight of software‑supply‑chain practices in the fast‑moving generative‑AI market. The episode underscores the need for rigorous verification of AI tooling before it reaches production environments.
Sources
Back to AIPULSEN