Openai announced the launch of the Bug Bounty program, which will allow researchers of cybersecurity to detect vulnerability in the products and receive payment For a message about them through the crowdsourcing platform bugcrowd .
The size of the remuneration depends on the level of danger and the impact of errors and varies from $ 200 for minor problems to $ 20,000 for “exceptional discoveries.”
However, despite the fact that the API Openai and ChatGPT are goals for bughangers, the company asked researchers to report on the problems of the model through a separate form if they do not affect safety. This solution is due to the fact that the problems of the model of the model do not fit into the Bug Bounty program, since they are not separate errors that can be corrected directly
Other possible problems include jailbreaks and safety circuit breeds that ChatGPT users use to force the chatbot to ignore Openai safety measures.