Misbehavior1 sourceFirst seen
DeepSeek's Chatbot Hacked Itself, Then Offered Up the Login Token
Ourtake
Prompt injection still has no fix, so the chatbot happily wrote its own XSS attack. DeepSeek patched it in a day.
What happened
A security researcher showed that DeepSeek's chat app could be prompt-injected into producing a cross-site scripting payload. In a proof of concept, the payload could steal a user's session token from browser storage. The researcher reported it through the app's Contact Us feature, and DeepSeek fixed it about a day later.