OpenAI Now Pays Up to $172,000 for a Single ChatGPT Bug

OpenAI is raising the rewards in its Security Bug Bounty program, and the biggest finds in ChatGPT can now pay up to about $172,000 once bonuses are added. The company announced the change on Monday, Nov. 2.

OpenAI says it is more than quadrupling the midpoint of its reward ranges for medium, high, and critical findings, on average.

The new numbers

A critical vulnerability in a core product, which OpenAI defines as ChatGPT, Codex, or the API Platform, now earns up to $30,000. Exceptional critical findings on those products remain eligible for up to $100,000, and now up to $172,000 with bonuses.

OpenAI’s own illustrative examples show the jump. A zero-click ChatGPT account takeover that historically earned $6,500 would now pay $51,577 after a quality bonus, a priority campaign bonus, and a top loyalty tier bonus. A significant security boundary break, such as being able to view other ChatGPT users’ chat messages, would pay $171,925 under the same multipliers.

New bonuses

Three new multipliers stack on the base bounty: a quality bonus for reports with a reliable proof of concept, a 5-tier loyalty program for repeat researchers, and priority campaigns with dedicated bonus pools for newly launched features.

Backdated payouts

Reports that were already submitted but not yet awarded will get the new reward ranges. 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.

A security sprint with Cathedral

In a separate post, OpenAI said it is planning its next internal security sprint with outside partners, including security firm Cathedral, whose researchers will use OpenAI’s models alongside OpenAI’s teams to challenge its defenses. OpenAI says its last sprint dedicated 25% of its product engineers to finding and fixing weaknesses, using Codex to investigate findings and prepare patches.

OpenAI’s security team will also be at the PROMPT2PWN event in Oakland on Nov. 20 and 21, where researchers can test targets in person.