# GitHub Is Restructuring Its Bug Bounty for a World Where AI Finds the Bugs First
The math used to be simple: find a critical vulnerability in GitHub, collect $20,000 to $30,000. Starting July 27, that ceiling drops to a flat $10,000 — and if you want the old money, you'll need an invitation.
GitHub announced this week that it's slashing public bug bounty payouts by roughly half across every severity tier and reserving its top rewards for a permanent, invite-only VIP program. The restructuring isn't just about money. It's a signal that the economics of vulnerability research are shifting faster than most bug hunters expected, and that the platform companies are adjusting their programs accordingly.
## The New Math, By Severity
The changes are stark on paper. Under the current public program, a high-severity finding can fetch between $10,000 and $20,000. After July 27, it pays a fixed $5,000. Critical drops from a range of $20,000-$30,000-plus to a flat $10,000. Low-severity reports fall hardest on a percentage basis — from a floor of $617 down to $250.
GitHub frames the fixed rates as a feature, not a concession. "You don't earn more by submitting more," the company wrote in its announcement. "You earn more by submitting better." Fixed amounts, the argument goes, eliminate uncertainty and reduce the triage burden that comes with evaluating borderline reports.
The VIP tier tells the other half of the story. Invite-only researchers will earn $1,000 for low severity, $7,500 for medium, $20,000 for high, and $30,000 or more for critical. That's the old public ceiling, moved behind a velvet rope. Qualification requires at least one critical, two high, four medium, or seven low-severity findings — though GitHub hasn't specified a time window for hitting those thresholds, and the announcement doesn't guarantee that qualifying researchers will actually receive an invitation.
Reports submitted before July 27, including those sitting in GitHub's triage queue, keep the old payout terms.
## Noise Reduction, or Talent Suppression?
There's a legitimate case for what GitHub is doing. Bug bounty programs at major platforms attract a long tail of low-quality, duplicate, and speculative submissions that consume security engineering time without producing actionable findings. Cutting public rates and raising the bar for VIP access concentrates interaction with researchers who have already demonstrated they can find real issues.
The HackerOne Signal threshold GitHub mentioned — but declined to specify — is the mechanism. Researchers below it get up to four initial submissions before their reports are throttled. HackerOne's general rules already cap new researchers at four trial reports per program in a rolling 30-day window, so the overlap isn't incidental.
But the structure also squeezes out researchers who aren't yet known quantities. A first-time finder who hits a critical GitHub vulnerability on their inaugural submission gets $10,000 and a path to VIP consideration. The same researcher who submits seven low-severity findings to earn their way in has spent significant time for $1,750 total at new public rates, with no guarantee of an invitation at the end. The program is being designed around researchers GitHub already trusts, not researchers who are trying to earn that trust.
## The AI Timing Isn't Coincidental
GitHub's announcement landed one day after Google introduced Gemini 3.5 Flash Cyber, a lightweight model fine-tuned to find, validate, and patch software vulnerabilities. In Google's internal testing, the model found 55 unique confirmed vulnerabilities in V8, compared with 47 for standard Gemini 3.5 Flash and 36 for Claude Opus 4.6. A Google Cloud team used it to find remote code execution flaws in public APIs and a memory-corruption issue in a production service — within two hours. The model then generated what Google called a 100%-reliable RCE exploit that bypassed ASLR and W^X protections.
Those benchmark figures are Google-reported and haven't been independently verified. But even discounted, they point at something real: AI vulnerability research is getting cheaper, faster, and more capable at scale. Models like Gemini 3.5 Flash Cyber can scan more code paths per hour than any human team, at a fraction of the cost.
For GitHub — which is owned by Microsoft and runs Copilot's code infrastructure — this isn't an abstract trend. Internal security teams can give an AI agent repository context, project-specific threat models, and a validation environment tuned to the actual running system. They can then test findings, generate proofs of concept, and push fixes before any external researcher sees the vulnerable code in a release.
When your own AI finds the bug first, external reports don't provide new information. They provide noise.
## What Changes for Independent Researchers
The practical effect depends entirely on where a researcher sits in the ecosystem.
Established, high-signal researchers get better outcomes: faster responses, higher ceiling, closer access to GitHub's security engineering team. For them, the VIP tier is a genuine upgrade. For everyone else, the public program is less attractive by design.
The researchers most affected are competent but not yet credentialed — people building a track record, people who have found real issues before but at lower severity, people from regions where bug bounty income is meaningful. A 50% pay cut across severity levels, combined with submission throttling for new participants, narrows the pipeline of researchers who can afford to develop expertise against GitHub-scale targets.
Whether that narrows the attack surface GitHub faces — or just narrows who gets paid for finding it — is a different question. Vulnerabilities don't care who the program rewards.
---
## HackWire Analysis
GitHub's restructuring reflects a calculation that's becoming visible across the industry: as AI tools commoditize vulnerability discovery at scale, the value of an individual external report decreases. What remains valuable is a trusted researcher relationship — someone who knows how to validate, prioritize, and communicate findings in ways that map to an engineering team's priorities.
That's the logic behind the VIP tier. GitHub isn't paying for bugs. It's paying for researchers who behave like an extension of the security team.
The risk in this model is that it optimizes for known knowns. The researchers with established Signal scores and multiple confirmed criticals are, by definition, working in areas the program has already validated. The researcher who finds something genuinely novel — in a feature or infrastructure component that hasn't been heavily bounty-hunted — may be working without an established reputation in exactly that area.
The Google Gemini announcement deserves more scrutiny than it received in most coverage. A lightweight model that can generate a 100%-reliable RCE exploit bypassing modern mitigations — even in a controlled environment against Google's own infrastructure, even as a demonstration — is a capability shift. Defenders should read that less as "Google's internal AI is impressive" and more as a preview of what motivated external actors can build with access to similar fine-tuning pipelines.
For security teams at companies running significant bug bounty programs: the pressure to demonstrate AI-augmented internal scanning is going to intensify. Bug hunters who can document AI-assisted discovery workflows may find that capability more valued than raw finding count. And for independent researchers — especially those early in their careers — programs with more accessible public tiers and transparent qualification criteria are about to look more attractive than they did last week.
— HackWire Editorial
---
## Related Coverage