Well documentedThree or more independent publishers report the core facts.
Slack AI Assistant Vulnerability Exposes Private Channel Data
Security researchers reportedly discovered a vulnerability in Slack's AI assistant that could be exploited through prompt injection techniques to leak sensitive information and access data from private channels without authorization.
19 August 2024 · Slack · Security and cyberattacks
Security researchers uncovered a vulnerability in Slack's AI assistant that could be exploited to leak sensitive information and unauthorized access to private channel data.
The AI assistant, introduced in September 2023 to summarize messages, answer questions, and search files, could reportedly be tricked via prompt injection techniques into sharing secrets with unauthorized users.
Reported impact
The vulnerability could potentially expose workplace conversations and sensitive data from private channels to unauthorized parties.
Index analysis
The flaw involved prompt injection, a technique where attackers manipulate AI system inputs to bypass safety controls and access restricted information.