A security flaw in the Cursor AI editor allows malicious commands to execute on users' PCs, particularly affecting Vibe Coding tasks.
A significant security flaw has been identified in the Cursor AI code editor, putting its users at risk of remote command execution when opening project folders. This vulnerability primarily impacts those engaged in Vibe Coding, where programming tasks are managed by AI models.
According to SlowMist, the risk arises when users utilize the "Open Folder" function. If a project folder is structured maliciously, it can trigger the automatic execution of harmful commands on both Windows and macOS systems. Several Cursor AI users have reportedly fallen victim to this exploit, although the full extent of the damage is still unclear.
Cos, the founder of SlowMist, has alerted the Cursor security team about the issue, sharing details and proof of concept to expedite resolution. As of now, Cursor has not issued a public response regarding the vulnerability.
In light of this threat, web3 researcher DeFi Teddy has recommended that users keep their Vibe Coding activities separate from devices used for cryptocurrency storage. He emphasized the importance of avoiding project downloads from unverified sources, cautioning that high-privilege programs like Cursor can potentially control users' computers.
This vulnerability comes on the heels of another incident reported by Oasis Security, which discovered a similar flaw in a different program that allowed for the injection of malicious code without user interaction. Cursor, developed by Anysphere, is gaining traction among developers, boasting around one million users who collectively produce over a billion lines of code daily. The company recently secured $900 million in funding, raising its valuation to $9 billion.
