Dynamic Watch Beating AI News: Cursor has released its findings into the此前的 Claude Code plugin poisoning incident. After a two-week investigation, the company discovered that the attack originated from its AI service provider, Fireworks. According to Cursor, Fireworks leaked an access key granted by Cursor. Hackers used this key to copy historical Coding Agent session records stored in Cursor’s shared storage. These records primarily came from user sessions in 2025 that had privacy mode disabled and were originally intended for model training. More seriously, some of these records contained API keys that had not been sanitized. Leveraging these credentials, the attackers further infiltrated certain internal software services used by Cursor, gaining access to archived non-production code, customer support emails, product usage data, partial AI conversation summaries, and credentials previously submitted by users to customer support. On October 29, the hackers used credentials stolen from within Cursor to directly tamper with the extension store’s cache, distributing malicious versions of Claude Code 2.1.312 and 2.1.313 containing SSH backdoors to users. The malicious versions were available for approximately 11 hours and may have been downloaded by up to 5,000 users. The upstream plugin marketplace Open VSX was not compromised, and Anthropic’s official versions remained unaffected. Cursor disclosed that the compromised Coding Agent session records affected fewer than 0.05% of users, with fewer than 0.01% of users having their API credentials exposed. Sessions with privacy mode enabled, customer code repositories, and production systems were not impacted by this attack. Cursor stated that the last detected malicious activity occurred on November 1, when it was successfully intercepted. The company has since notified affected users, completely ceased using Fireworks for inference requests, revoked its data access permissions, rotated internal credentials, and redeployed its production infrastructure.
Cursor Discloses That the Hack Originated from a Fireworks API Key Leak
MarsBitShare
Summary
Cursor has disclosed that a recent crypto hack originated from a leaked Fireworks API key, which allowed attackers to access historical coding agent records stored in shared space. These records, dated 2025 and not in privacy mode, contained unmasked API keys used for model training. The attackers exploited these credentials to infiltrate internal services and distribute malicious versions of the Claude Code plugin. The exchange hack impacted up to 5,000 users for 11 hours. Cursor has since revoked access, rotated credentials, and redeployed its infrastructure to contain the breach.
Source:Show original
Disclaimer: The information on this page may have been obtained from third parties and does not necessarily reflect the views or opinions of KuCoin. This content is provided for general informational purposes only, without any representation or warranty of any kind, nor shall it be construed as financial or investment advice. KuCoin shall not be liable for any errors or omissions, or for any outcomes resulting from the use of this information.
Investments in digital assets can be risky. Please carefully evaluate the risks of a product and your risk tolerance based on your own financial circumstances. For more information, please refer to our Terms of Use and Risk Disclosure.