Attackers Backdoor Official Anthropic Claude Extension on Cursor Marketplace, Exposing Up to 5,000 Developers
A malicious version (2.1.313) of Anthropic's official Claude VS Code extension was distributed through Anthropic's verified Cursor account for nearly 12 hours, installing a persistent SSH backdoor and harvesting environment data on up to 5,000 developer machines. Cursor removed the package, cut off the attacker's server and audited its marketplace, but neither Cursor nor Anthropic has explained how the account was compromised, and Anthropic has not publicly warned customers.
Skynet Chance (+0.01%): The incident shows that the distribution channels of AI coding tools, which run with broad system permissions, can be hijacked by attackers. This is a conventional security failure rather than an AI alignment or loss-of-control issue, so it only marginally raises concern about the security of AI infrastructure.
Skynet Date (+0 days): A single human-driven supply-chain attack does not change the pace of AI capability development or of AI risk emergence.
AGI Progress (0%): The breach involves extension distribution security and says nothing about AI model capabilities or progress toward general intelligence.
AGI Date (+0 days): The incident is unlikely to affect research funding, compute or development speed, so it has no impact on the AGI timeline.
<< All AI news for October 30, 2026
[ Get the daily index digest on Telegram → ]Related AI News
- Mistral Open-Sources Weights of Trillion-Parameter Mistral Large 4, Pitching It for Cyber Defense 2026-10-29
- Wiz CTO Warns AI Is Accelerating Vulnerability Discovery After Critical PHP Bugs 2026-10-28
- Wiz CTO Warns AI-Accelerated Vulnerability Discovery Threatens Aging Internet Code After PHP Bugs 2026-10-28
- Google Unveils Gemini 4 Carbon With Strong Coding and Cybersecurity Gains, Withholds Public Release 2026-10-28
- Google DeepMind Unveils Gemini 4 Carbon with Leading Cybersecurity and Multi-Day Agentic Capabilities 2026-10-28