NOVEMBER 16, 2026
Subscribe
Global Press Media · World Report
Technology

Stolen Access Key Connects Fireworks AI Hack to Compromise of Cursor’s Claude Code Extension

Stolen Access Key Connects Fireworks AI Hack to Compromise of Cursor’s Claude Code Extension

Security researchers have linked a recent compromise of Anthropic's Claude Code extension on Cursor's marketplace to a stolen access key that emerged during a separate breach of Fireworks AI. The incident exposed internal credentials, enabling an attacker to acquire a key that Cursor had previously issued to the Fireworks platform for integration purposes.

In a statement released by Cursor, the malicious party is said to have employed the illicit key to breach the Claude Code extension, a utility that lets developers generate and edit code using Anthropic's AI model directly inside the Cursor development environment. The statement reads, "An attacker obtained an access key we had issued to Fireworks," confirming the connection between the two events.

The episode underscores the intertwined risk profile of AI tooling ecosystems, where shared authentication mechanisms can become single points of failure. Anthropic, the creator of Claude, has not disclosed the full scope of data exposure but warned that the compromised extension might have permitted unauthorized access to user prompts and code snippets processed through the service.

Cursor, a developer‑focused platform that hosts a marketplace for extensions, has temporarily withdrawn the Claude Code add‑on while conducting a forensic review. The company says it is revoking the compromised key, rotating all credentials tied to third‑party integrations, and tightening key‑distribution policies to avert similar incidents.

Fireworks AI, a startup delivering generative visual‑content tools, suffered a breach earlier this month that revealed internal systems and several API keys. Although the breach did not directly target Cursor, the leaked key appears to have been repurposed by threat actors to gain a foothold in the broader AI extension ecosystem.

Industry analysts point out that the situation highlights the necessity of strong secret‑management practices, especially as AI services become more modular and depend on cross‑platform authentication. Experts advise the use of short‑lived tokens, zero‑trust network architectures, and continuous monitoring of credential usage to mitigate fallout from inevitable future leaks.

Both Anthropic and Cursor have pledged to inform affected users and provide guidance on securing their accounts. As investigations proceed, the incident serves as a reminder that the rapid growth of AI‑driven developer tools brings heightened security responsibilities for providers and users alike.

Source: feedburner
Editorial Desk — Editorial desk.

Comments (0)

Be the first to comment.

Join the discussion

Protected by reCAPTCHA v3

Related