Osegi
Security·1 min read

OpenAI Raises Bug Bounty Rewards to as Much as $172,000

OpenAI has overhauled its vulnerability reporting program for ChatGPT, Codex and its API, with rewards reaching about $172,000 in exceptional cases.

OpenAI logo beside a padlock and lines of code, symbolizing the expanded bug bounty program for security researchers.
Image: اليوم السابع

OpenAI has raised the rewards it pays security researchers who find vulnerabilities in its products. In exceptional cases, payouts can reach about $172,000 once bonus incentives are counted. The updated program covers ChatGPT, Codex and the API platform.

According to the announcement on the company's website, OpenAI has increased the median reward ceilings for medium, high and critical vulnerabilities by more than fourfold. The base reward for some critical flaws in its main products now reaches $30,000. Exceptional cases remain eligible for a base reward of up to $100,000 before any bonuses are added.

Report quality drives the payout

The update rewards the quality of a submission, not just the discovery of a problem. OpenAI grants extra incentives to reports that include practical proof the flaw can be exploited, and that explain the conditions needed for it to occur and the resulting impact.

The company also introduced a five-tier loyalty system for researchers who work with it on an ongoing basis. It added targeted campaigns as well, aimed at testing new features in the early stages of their release.

Retroactive payments

The new reward ranges apply to earlier reports that have not yet been paid out. OpenAI will also pay the difference owed to some researchers who received awards in the 30 days before the announcement.

The change reflects an effort to widen the role of independent researchers in finding problems before they can be exploited. It also draws a line between base reward limits and the top figures, which are only reached when the extra incentive conditions are met. Not every researcher who finds a critical flaw will automatically receive the advertised maximum.

Apps worth a look

SUMO: the number puzzle

Tap the right tiles, add them up, hit the target. Easy to learn, hard to master. Free, for all ages.

Tags:openaibug bountychatgptcodexvulnerabilitiesapi

More from this day

See all →

More in: Security

See all →