MisbehaviorSerious1 sourceFirst seen
Slack AI Gets Tricked Into Leaking Your API Key Through a Channel Nobody Reads
Ourtake
Public channels nobody remembers creating are now an attack surface for your private secrets.
What happened
Security researchers at PromptArmor disclosed that Slack AI could be manipulated through indirect prompt injection planted in a public channel, causing it to render a link that leaked a user's private API key to an attacker. Slack reviewed the report, called the evidence insufficient and described the behavior as intended; PromptArmor published publicly so users could adjust settings.