OpenAI Raises Bug Bounty Rewards, With Critical Finds Paying Up to $172,000

Bigger base rewards for serious vulnerabilities

According to OpenAI, the update is meant to encourage independent researchers to look for vulnerabilities in its software that the company's own testing may have missed. The company describes responsible disclosure as a core part of its security model, one that complements its Defense Factory work, where engineers use OpenAI models to find weaknesses, reproduce them and test fixes.

A critical vulnerability in a core product such as ChatGPT, Codex or the API Platform will now earn up to $30,000. Exceptional critical findings on those targets remain eligible for up to $100,000, and now up to $172,000 with bonuses. The full terms are published in the program's own documentation.

Three new ways to earn more

OpenAI is adding several bonus mechanisms on top of the base bounty.

• Quality multiplier: High-quality reports can earn a bonus on the base amount. OpenAI points to a reliable proof of concept that isolates the bug, shows when it can be exploited and demonstrates its impact as the kind of work that gives its team a head start on a fix.
• Loyalty bonuses: A new 5-tier system provides scaling bonuses for researchers who consistently report to OpenAI.
• Priority campaigns: Short-term campaigns will offer dedicated bonus pools for newly launched features, giving researchers early, high-reward chances to test new developments.

How the payouts add up

OpenAI published illustrative examples, noting that all scenarios and historical awards are illustrative. A medium-severity ChatGPT issue, where a read-only member can upload files to a Project without write permission, previously earned $750. Under the new scheme it is calculated as $2,000 × 1.15 quality bonus × 1.05 Bronze loyalty tier bonus, for $2,415.

A high-severity IDOR in Prism, a non-core asset, that allows reading a file uploaded to a user's Project previously earned $1,000. It is now $2,500 × 1.15 × 1.10 Gold loyalty tier bonus, or $3,162.50.

For a critical one-click account takeover in ChatGPT, the historical award was $6,500. The new calculation is $30,000 × 1.15 quality bonus × 1.30 priority campaign bonus × 1.15 for the 31337 loyalty tier, giving $51,577. An exceptional finding, such as the ability to view arbitrary ChatGPT users' chat messages, was previously $100,000 and comes to $171,925 under the same multipliers.

Retroactive payments and an in-person event

Researchers who have already submitted a report that has not yet received a bounty will have the new reward ranges applied. For bounties awarded in the 30 days before the announcement, OpenAI says it will pay the difference if the finding would have earned more under the new ranges, covering the base award and loyalty awards.

OpenAI's security team will also attend PROMPT2PWN in Oakland on November 20 to 21, where researchers can test targets in person. The company says it wants to hear what researchers are working on and how to partner with them.

For security researchers, the changes mean higher potential payouts, particularly for well-documented reports and for those who report repeatedly. OpenAI says it wants to be "a team you come back to when you find something we've missed."