Atlassian closes one-click Rovo data leak flaw

Atlassian has patched a security vulnerability known as RovoBlast in its artificial intelligence assistant, Rovo. The flaw allowed attackers to steal sensitive corporate data from a single malicious link by tricking authenticated users into clicking the link. The Rovo Chat feature processed external information embedded in URLs, making it possible for an attacker to create a link with malicious instructions.

These instructions would enter the victim's Rovo session and direct the AI assistant to locate and transmit information to the attacker's infrastructure. The vulnerability was exploited by manipulating a Rovo URL parameter called "rovoChatPrompt," which was designed to pre-populate the assistant's chat interface. Rovo operated with the legitimate access rights of the victim, meaning it could not automatically access information beyond the employee's permissions, but it could still extract data based on the compromised AI session's authorization.

The flaw was reported through Atlassian's vulnerability disclosure process and was classified as a P2 security issue. Atlassian released a server-side fix on July 8 and validated the remediation. Employees do not need to install a separate patch since the correction was made within Atlassian's cloud infrastructure. No CVE identifier had been assigned as of August 8, and there was no evidence of exploitation outside controlled security testing.

The discovery of RovoBlast highlights the risks of enterprise AI assistants that combine broad data access with autonomous actions, as traditional permission controls may not prevent manipulated AI agents from redirecting authorized information.