OpenAI pays users of the AI chatbot ChatGPT up to $20,000 to find bugs and vulnerabilities in the system. In a blog post, the company announced that it would start a program to find vulnerabilities in its AI products. In order to communicate more effectively on vulnerability detection issues, OpenAI has entered into an agreement with the platform Bugcrowd Inc.which collects requests from program participants.
OpenAI is willing to pay starting at $200 for vulnerability discovery “low severity” and up to $20,000. “for extraordinary finds”. According to the company, it decided to launch the program because it believes in it “Transparency and collaboration [с пользователями]» are key to identifying problems that a particular technology may have.
“This initiative is a natural part of our overall vision to develop safe and advanced AI. We want to create safe and reliable technologies and services, so we would like to ask for your help.”according to the company’s blog.
On the Bugcrowd platform, the OpenAI Vulnerability Program topic identifies a set of security issues for which no bounty is awarded. In particular, this includes requests to ChatGPT related to extorting prohibited information from the bot by deception, writing malicious code or requests where the user forces the chatbot to write insults.