OpenAI is popping to the general public to search out bugs in ChatGPT, asserting a “Bug Bounty Program” to reward individuals who report any safety flaws, vulnerabilities or different points inside the AI system.
The bounty is open to anybody from precise researchers to normal individuals who identical to exploring know-how. Rewards come within the type of money prizes with “low-severity findings” beginning at $200 and “distinctive discoveries” going all the best way as much as $20,000. Bugcrowd, a bug bounty platform, is dealing with submissions and payouts.
Google and Apple are among the many tech corporations which have beforehand carried out bug bounty packages. In 2019, Google paid out $6.5 million to individuals who reported points, giving as a lot as $201,337 in a single reward. Apple went even additional to supply as much as $2 million for anybody that finds “points that bypass the precise protections of Lockdown Mode.”
ChatGPT has struggled with bugs — final month your entire system went offline after customers reported seeing names of conversations they weren’t part of. Then, a number of days later, a Twitter user posted that that they had discovered greater than 80 secret plugins whereas hacking ChatGPT.
Not all points reported to OpenAI will warrant a money prize, together with jailbreaking or getting the mannequin to say or faux to do something adverse. The corporate’s bug bounty announcement tries arduous to point out it cares about privateness and safety, but additionally provides, “Whereas we work arduous to stop dangers, we won’t predict each means individuals will use or misuse our know-how in the actual world.” Time will inform if this initiative will do something to stop it.