Rovo AI Vulnerability Exposes Enterprise Data

Sunday, 22 November 2026
Latest News · In-Depth Guides · Information & Tips to Keep You Safe

← Back to articles

Rovo AI Vulnerability Exposes Enterprise Data

Security researchers at Varonis have disclosed a flaw in Rovo, Atlassian’s enterprise AI assistant, that could have let an attacker steal company data with a single click on a specially crafted link. SecurityWeek reports that Atlassian fixed the issue before the findings were published.

What the flaw was

The vulnerability, which Varonis named RovoBlast, was presented at the DEF CON conference. According to the report, a link could seed attacker-controlled instructions directly into a user’s live AI session. It required no jailbreak and no permission bypass, because the assistant treated externally supplied parameters as trusted input. The link used a URL parameter called rovoChatPrompt, which pre-fills Rovo’s chat window. Varonis calls this kind of attack parameter-to-prompt (P2P) injection and says it previously reported a similar issue in Microsoft Copilot called Reprompt.

The researchers also found that the organization ID in the link could be left blank, and Atlassian would still route the request into the victim’s own default organization, with no sign that the session had been seeded from outside.

What could have been exposed

Rovo works across Jira, Confluence, Bitbucket and connected tools such as Slack, Microsoft 365 and Google Workspace, and has agent features that can complete multi-step tasks. When the researchers asked Rovo what data it could see, it listed a wide range of sources. The actual leak came from a built-in tool called ResearchAgent, which can run web research and move across arbitrary sites. Once the attacker’s instructions were seeded, that same ability let Rovo pull internal data and send it out to the open web in one automated chain. The team demonstrated this in three proof-of-concept cases, extracting Confluence pages, Jira tickets and SharePoint content containing personal data. In general, one seeded link was enough.

What the researchers recommend

  • Limit which systems the AI assistant can reach and disconnect unused integrations
  • Wall off sensitive areas such as legal, HR and finance
  • Turn off browsing or multi-step automation features that are not in active use
  • Monitor the assistant’s activity logs routinely

What it means for everyday users

The same lesson applies beyond Atlassian: a link can do more than open a page when an AI assistant is involved, so treat links from unknown senders with the same caution as unexpected attachments. Our guide on how to spot a phishing email quickly covers the checks.

Source: Read the original report

Related reading

Check it yourself with our free tools

  • Free Temporary Email Generator: Get a free temporary email address (10 minute mail) for sign-ups, trials and verification codes.
  • Password Breach Checker: Free password breach checker: find out if your password has been leaked in a data breach.
  • Scam Email Checker: Free scam email checker: paste a suspicious email or message and check it for phishing and fraud red flags.
Scroll to Top
Privacy Policy · Terms of Use · Cookie Policy