Declassifying the May 3rd, 2022 Responsible Disclosure of the Prompt Injection Attack Vulnerability of GPT-3

Last updated Sept 23, 2022

Prompt Injection has been in the news lately as a major vulnerability with the use of instruction-following NLP models for general purpose tasks. In the interest of establishing an accurate historical record of the vulnerability, we are de-restricting the previously secret Responsible Disclosure which Preamble (we are a small A.I. safety startup - made on May 3rd, 2022 to OpenAI.

May 3rd: the Discovery, and Immediate Responsible Disclosure

May 3rd: OpenAI Confirms Receipt of Disclosure

May 4th: Provided Additional Examples

May 19th: Details Provided for Approaches to Mitigate the Danger