The OpenAI bug bounty just got much more generous. As of November 3, 2026, OpenAI pays up to $30,000 for critical flaws in core products like ChatGPT, Codex and the API Platform, and up to about $172,000 for exceptional finds.
Quick Facts
- Announced November 2-3, 2026, alongside a planned security sprint.
- Midpoints of the medium, high and critical reward ranges are up more than 4x on average.
- Critical bugs in ChatGPT, Codex and the API Platform: up to $30,000.
- Exceptional critical finds: $100,000 base, up to roughly $172,000 with bonuses.
If you hunt bugs for a living or as a hobby, those numbers change the math on whether OpenAI’s products are worth your weekends.
What exactly changed in the OpenAI bug bounty?
OpenAI didn’t just add a bigger number at the top. According to Help Net Security, the midpoints of the medium, high and critical ranges all went up, by more than four times on average.
The top end now has two levels. Standard critical bugs in core products cap at $30,000. Exceptional critical findings start at a $100,000 base and can climb to about $172,000 once bonuses are added.
Here’s the new payout picture at a glance.
| What you find | What OpenAI will pay |
|---|---|
| Critical bug in a core product (ChatGPT, Codex, API Platform) | Up to $30,000 |
| Exceptional critical finding, base award | Up to $100,000 |
| Exceptional critical finding, with bonuses | Up to about $172,000 |
| Medium, high and critical ranges (midpoints) | More than 4x higher on average |
The table shows why the exceptional tier matters. It’s more than five times the standard critical cap.
How do bonuses push a payout to $172,000?
The base award is only part of the story. OpenAI says three things can raise it.
- A quality multiplier for reports with a reliable proof of concept that isolates the bug, shows when it can be exploited and demonstrates impact.
- A five-tier loyalty program that pays larger bonuses to researchers who keep working with OpenAI.
- Short-term campaigns with dedicated bonus pools for testing newly launched features.
In plain terms, a clean, well-documented report from a repeat researcher on a brand-new feature is the best-paid combination. A vague write-up of a minor issue is not.
Will you get paid more on reports you already filed?
Probably, yes. Reports that haven’t been awarded yet will be judged under the new ranges.
OpenAI is also looking backward. If you were paid in the 30 days before the announcement and the new ranges would have paid more, the company says it will cover the difference, including base and loyalty awards.
What is the security sprint behind it?
The bounty change came with a planned security sprint. Outside researchers from Cathedral Systems will work with OpenAI’s own teams, using OpenAI models to probe existing defenses, test fixes and keep watch over time, according to Cyber Security News.
No start date or target list has been published, and there are no results yet. Treat it as a plan, not a finished audit.
It fits a pattern we’ve been tracking. OpenAI recently used AI monitors after a supply-chain scare, which we covered in OpenAI’s AI monitors and the Hugging Face breach. Bigger rewards are the human side of the same push.
Why should everyday users care?
You’ll never file a bug report, but you still benefit. Flaws found by researchers get fixed before criminals use them. That matters more as AI tools hold our chats, code and API keys.
Recent incidents show the stakes. After a vendor breach, teams had to rotate their keys, as we explained in our Northflank breach guide. Better bounties are one way to catch problems earlier.
How do you write a report that earns the multiplier?
OpenAI says the multiplier goes to reports with a reliable proof of concept. That gives you a simple checklist for a strong submission.
- Show the bug working, step by step, so a triager can reproduce it without guessing.
- Explain when it can be exploited, including any conditions or accounts needed.
- Describe the real impact, such as what data or access an attacker would gain.
OpenAI’s own sprint write-up admits that duplicate findings and unclear severity slowed its triage, so a tidy report saves everyone time. It’s also the easiest way to stand out from the pile.
One more tip: the priority campaigns reward testing brand-new features. Newly launched products tend to have the least-tested code, so that’s where fresh bugs often hide.
Frequently Asked Questions
How much does the OpenAI bug bounty pay now?
Critical bugs in core products pay up to $30,000. Exceptional critical finds start at $100,000 and can reach about $172,000 with bonuses.
Which products are covered?
The top critical tier names ChatGPT, Codex and the API Platform. Check OpenAI’s own announcement for the full scope before you start testing.
Do I need a proof of concept?
You don’t strictly need one, but a reliable proof of concept earns a multiplier on the base bounty. Clear reports are paid better.
When does the security sprint start?
OpenAI hasn’t published dates. The announcement describes a planned effort with Cathedral Systems and no results so far.
Our take
This is a smart move, and an overdue one. When AI tools sit between people and their most sensitive data, paying researchers well is cheaper than cleaning up a breach. We’d still wait for the sprint’s actual results before calling OpenAI’s defenses stronger. Big payouts show intent; fixed bugs show progress.
Leave a Reply