OpenAI, the startup behind the favored ChatGPT AI author, has introduced the launch of a brand new bug bounty program with some fairly vital rewards for probably the most “distinctive discoveries.”
Money-based rewards are set to vary from $200 for low-severity findings to as a lot as $20,000, with members requested to deal with vulnerabilities, bugs, and safety flaws.
The corporate says it’s doing this to harness a extra clear and collaborative atmosphere which is a vital step in opening up the expertise amid speculations of potential giant language mannequin (LLM) misuse.
OpenAI bounty program
Safety researchers, moral hackers, and expertise fanatics are all being requested to come back collectively and assist OpenAI to search out – and perceive – its flaws. A devoted Bugcrowd web page (opens in new tab) has been set as much as deal with submissions and rewards.
Researchers are being requested to not submit mannequin questions of safety by way of the bug bounty program, and as an alternative to submit them by way of a separate type. OpenAI says it’s because investigating such points requires big quantities of analysis by specialists, and thus past the scope of the bounty program that provides as much as $20,000.
OpenAI explains: “Mannequin questions of safety don’t match effectively inside a bug bounty program, as they aren’t particular person, discrete bugs that may be immediately mounted.”
Nevertheless, different safety bugs pertaining to ChatGPT are throughout the scope for bounties, together with API targets, third-party company targets, OpenAI API keys, OpenAI Analysis Org, and different OpenAI targets. Every class has its personal tiers for reward paybacks, and never all are eligible for the complete $20,000.
A full and detailed record of all of the features each included and excluded from this system is obtainable on the Bugcrowd web page (opens in new tab).