#Open API Assistants exploitable through post/get prompt injection

1 messages · Page 1 of 1 (latest)

serene pivot
#

I don't want to educate the masses, but post get prompt injection even with good masking is possible. HIGHLY advise looking at this as it allows any Assistant API to have it's instructions modified on-the-fly to use for API calls under any prompt. no pics, and not educating anybody. This is serious.

serene pivot
#

Main page is exploitable for information and similar isues as well.

#

fulyl comprimises all account information for users and for openai code.

#

<@&1009772329649320030> this is also being reported through the standard channel with more information by another user.

#

Its a full api reverse