OpenAI more than quadruples critical bug bounties

OpenAI announced that it is planning its next internal security sprint. Outside researchers and security partners will join its own teams to test OpenAI's systems using OpenAI models.

The first named partner is Cathedral Systems. Its researchers will use OpenAI's models alongside internal staff to challenge the company's defenses. OpenAI says the firm brings "differentiated expertise and approaches" to finding weaknesses, but has not said how big the engagement is or how long it will run.

OpenAI is also changing its Security Bug Bounty program. In a follow-up post, it said it is more than quadrupling the midpoint of its reward ranges for medium, high and critical findings, on average. 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 remain eligible for up to $100,000, or up to $172,000 with bonuses. In one of OpenAI's examples, a one-click ChatGPT account takeover that previously paid $6,500 could now earn $51,577.

The bonuses include a multiplier for high-quality reports with a reliable proof of concept, a new five-tier loyalty program, and priority campaigns with dedicated bonus pools for newly launched features. Reports still awaiting a bounty will get the new ranges, and OpenAI will pay the difference on bounties awarded in the 30 days before the announcement if they would have earned more. Its security team will also be at PROMPT2PWN in Oakland on November 20–21, where researchers can test targets in person.

The new sprint builds on the previous one, when OpenAI put 25% of its product engineers on finding and fixing weaknesses. Engineers used Codex to investigate findings, reproduce vulnerabilities in isolated environments and prepare patches for review. OpenAI did not say how many vulnerabilities that effort found or fixed.

The company says "traditional cyber defenses alone are no longer sufficient," pointing to recent incidents across the industry. It argues that organizations need frontier cyber capabilities to keep "the defender's window open."

OpenAI also restated a commitment it had already made: $1 billion in subsidized access over six months for frontline defenders protecting essential services and digital infrastructure. It says its most capable models are available to Daybreak customers, and that it is sharing the architecture and lessons from the last sprint along with a blueprint for supporting cyber defenders in industry and government.

OpenAI has not said when the new sprint will start.