OpenAI raised the rewards it pays security researchers who report flaws in ChatGPT, Codex and the API Platform. An exceptional critical finding can now earn a base bounty of up to $100,000, and up to $172,000 once bonuses are added. Reward midpoints for medium, high and critical findings more than quadrupled on average, according to the company’s announcement as summarized by Help Net Security.
The day in tech, in one email: TLDR is the daily newsletter with links and TLDRs of the most interesting stories in startups, tech and programming.
We may earn a commission if you sign up.
Ordinary critical vulnerabilities in those core products carry a base bounty of up to $30,000. The extra money comes from a multiplier for high-quality reports and from a loyalty program with five tiers. Short campaigns add dedicated bonus pools for researchers who test newly launched features.
How $30,000, $100,000 and $172,000 fit together
The structure in Help Net Security’s Nov. 3 report on the changes has layers. A critical bug in ChatGPT, Codex or the API Platform sets the base. Exceptional criticals lift that base to $100,000. A report quality multiplier and loyalty-tier bonuses then stack on top, which is how the ceiling reaches $172,000. The payout examples OpenAI published appear only as an image in that coverage, so the exact multiplier values are not reproduced here.
OpenAI explained why quality matters to its engineers: a reliable proof of concept isolates the bug, shows when it can be exploited and demonstrates its impact, which speeds up a fix. The company also applied the new numbers backward. Reports that had been submitted without a bounty yet will be assessed under the new ranges, and for awards made in the 30 days before the announcement, OpenAI said it will pay the difference whenever the finding would have earned more under the new schedule, counting base and loyalty awards.
For researchers, the retroactive clause changes the value of work already done. Anyone with a pending, unawarded report does not need to resubmit, since OpenAI said it will apply the new ranges, and recent awardees may be owed a top-up. The company has not said how many people that covers or what the combined cost of the true-up will be, so the size of the change in actual payouts is not yet visible in the coverage.
Three steps from $20,000: 2023, 2025 and now
The program began on April 11, 2023, when OpenAI opened a bounty run through Bugcrowd with rewards starting at $200 for low-severity findings and topping out at $20,000 for exceptional discoveries, as described on OpenAI’s launch page for the program. Submissions and payouts are managed on the Bugcrowd engagement page for OpenAI.
In March 2025 the ceiling jumped fivefold. BleepingComputer’s March 28, 2025 report put the new maximum at $100,000 for what OpenAI called exceptional and differentiated critical findings, and noted that model-safety issues, jailbreaks and safety bypasses used by ChatGPT users stayed outside the scope. SecurityWeek reported the same move and a partnership with the startup SpecterOps for continuous adversarial red teaming. Google had announced a fivefold increase in its own bounty rewards in 2024, TechRadar noted at the time.
Promotions have been part of the program for a while. BleepingComputer reported that in 2025 OpenAI doubled payouts for insecure direct object reference vulnerabilities in its infrastructure and products until April 30, capped at $13,000, and the program ran through Bugcrowd from its start. The new short-term campaigns with dedicated bonus pools for newly launched features follow the same logic: pay more where fresh code has had the least scrutiny.
The arithmetic shows how far the program has moved. A $172,000 maximum is 8.6 times the $20,000 ceiling set in 2023 and 72 percent above the $100,000 base reached in 2025. The top figure applies only to exceptional criticals in the core products, while the broader change is the more-than-fourfold average rise in midpoints for medium, high and critical ranges, which affects far more of the reports a typical researcher files.
Defense Factory and the bounty’s place in OpenAI’s security work
The bounty program complements what OpenAI calls Defense Factory, in which its engineers use the company’s own models to find and reproduce vulnerabilities and to test fixes. TechRadar’s March 2025 coverage placed the earlier increase alongside an expanded Cybersecurity Grant Program, which had funded 28 research initiatives by then.
OpenAI has framed the spending as part of a wider security push. In 2025 the company described its commitment to rewarding meaningful, high-impact security research, and said it shares tradecraft with other AI labs to strengthen collective defenses, as TechRadar relayed. The Cybersecurity Grant Program it launched in 2023 has reviewed thousands of applications, funding topics from autonomous cyber defenses to prompt injection and secure code generation.
Two things remain unsettled in the public coverage. The full tier table behind the $172,000 maximum has not been published in text form, and whether the 2025 exclusions for model-safety issues and jailbreaks still apply to the new schedule is not stated in the summary of the announcement.
This article was produced with the assistance of AI and reviewed by Morning Overview editors prior to publication.
The day in tech, in one email: Get TLDR →
We may earn a commission if you sign up.
More from Morning Overview
- Long use of a common prostate pill is tied to a higher chance of glaucoma
- Ford is recalling 223,472 F-150 pickups because the fuel tank can leak or detach
- Hybrids have 15% fewer problems than gas cars, while EVs and plug-in hybrids have about 80% more, Consumer Reports finds
- Regulators cleared the first U.S. small modular reactor, 4 months early
