GitHub Bug Bounty Payout Cuts: New VIP Tier Explained

Effective July 27, 2026, GitHub is cutting the payouts on its public bug bounty program by at least half at every severity level, and shifting its largest rewards to a permanent, invite-only VIP tier. The Microsoft-owned company framed the change as a response to a problem now common across the vulnerability-disclosure industry: a rising tide of low-effort and AI-generated submissions that consume triage time without producing real security value.

The new economics are blunt. Under the revised public schedule, a critical finding that previously paid $20,000 to $30,000 or more now carries a fixed reward of $10,000. High-severity reports cap at $5,000, medium at $2,000, and low at $250. The top-end money doesn’t disappear — it moves. Researchers admitted to the invite-only VIP tier can earn $30,000 or more for critical work, along with faster responses and closer access to GitHub’s security engineering team. Reports filed before July 27, including those already sitting in GitHub’s triage queue, retain the previous payout terms.

The reasoning: “you don’t earn more by submitting more”

GitHub’s message to researchers was direct about the incentive it wants to change. “You don’t earn more by submitting more,” the company said. “You earn more by submitting better.” The restructuring, in GitHub’s telling, is meant to reduce noise while rewarding the smaller group of researchers who have repeatedly demonstrated they can find genuine, exploitable flaws.

The backdrop is a triage queue buried under volume. Like many large programs, GitHub’s has been strained by a surge of submissions that pattern-match to real bugs but don’t hold up — plausible-looking reports, often assembled with the help of large language models, that describe a vulnerability that isn’t actually there or can’t be exploited as claimed. Every one still has to be read, reproduced, and dispositioned by a human. As the ratio of noise to signal climbs, paying per accepted report starts to reward quantity in a way that works against the program’s purpose.

The move is not GitHub’s first step in this direction. It follows a May 2026 policy change that raised the bar for what a submission must include: a working proof of concept, demonstrated real-world impact, validation before filing, and closer adherence to program scope and the list of ineligible findings. The July payout restructuring extends that logic from what you must submit to how you get paid for it — pairing a higher evidentiary bar with a reward structure that concentrates money on proven performers.

The AI-generated report problem

The core pressure GitHub is reacting to is not unique to GitHub. Across the security industry, the cost of producing a vulnerability report has collapsed while the cost of evaluating one has not. A researcher can now point a model at a codebase or a disclosed CVE and generate a fluent, technical-sounding write-up in minutes. When the underlying analysis is sound, that’s a productivity gain. When it isn’t, the result is a report that looks exactly like real research until a human spends an hour proving it wrong.

That asymmetry is what breaks a pay-per-report model. Triage capacity is fixed and expensive; report generation is nearly free and effectively unlimited. Left unchanged, the incentives reward the researchers filing the most reports, not the ones filing the best — and the maintainers on the receiving end drown. GitHub is far from alone in feeling it: open-source projects and commercial programs alike have publicly complained about AI-assisted “slop” reports overwhelming their inboxes over the past year.

GitHub sits at a particularly sensitive point in that ecosystem. The platform is both a target worth attacking and the distribution layer for a huge share of the world’s software, which means the quality of its own security response has downstream effects. It’s also, increasingly, a venue where AI runs directly against code — the same surface that produced incidents like the GitLost prompt-injection flaw in agentic workflows, where automation acting on repository contents created a novel attack path. The tooling that makes it cheap to generate bug reports is the same class of tooling reshaping how the platform is used and attacked.

What it changes for researchers

For the top tier of bug hunters, the practical effect may be positive. The VIP program promises higher ceilings, faster turnaround, and a direct line to GitHub’s engineers — the things experienced researchers most often say they want, and rarely get, from crowded public programs. For a proven specialist, a program that pays $30,000-plus for a critical finding and actually responds quickly can be more attractive than one that nominally pays more but leaves reports languishing in a queue for months.

For everyone else, the calculus gets harder. A newer or occasional researcher now faces materially lower public payouts and a gate — the VIP tier is invite-only — standing between them and the top rewards. GitHub’s implicit bet is that the researchers capable of finding critical bugs will earn their way into the tier through track record, and that the ones deterred were mostly filing noise anyway. The risk is that a halved public bounty also discourages a meaningful number of legitimate part-time contributors, narrowing the funnel of talent that feeds any healthy disclosure program.

There is a broader industry signal here too. Bug bounties have spent a decade selling the promise that anyone, anywhere, could turn security skill into income by reporting flaws to the companies that own them. GitHub’s restructuring nudges that model toward something more selective and relationship-driven — fewer, more-trusted researchers, paid better, held to a higher bar. If a platform of GitHub’s scale can make that shift stick without losing coverage, expect others wrestling with the same AI-driven flood to study it closely. The economics of adversarial security research — where the supply chain of trust depends on flaws being found and reported responsibly before they’re exploited — are being rewritten in real time.

What it means

GitHub is trying to solve a genuine problem, and the direction it has chosen is defensible: when the marginal report is nearly free to produce and expensive to evaluate, paying per report stops aligning incentives with outcomes. Concentrating rewards on demonstrated performers is a rational answer to a signal-to-noise crisis that is only going to intensify as generative tools improve.

Who wins. Established, high-skill researchers with a track record — they get more money, faster responses, and closer access. GitHub’s security team wins too, at least in theory, trading a flood of low-value reports for a narrower stream of higher-quality ones it can actually act on.

Who loses. Newer and part-time researchers, who face lower public payouts and a locked door to the top tier. And, potentially, GitHub itself, if the halved public bounty deters enough legitimate contributors that real vulnerabilities go unreported — the exact outcome the program exists to prevent. The line between “noise” and “an inexperienced researcher who would have gotten good” is not always clear at triage time.

What to watch next. Three things. First, whether other major programs follow — a payout cut framed around AI slop, from a platform GitHub’s size, is the kind of precedent that spreads fast if it works. Second, GitHub’s own disclosure cadence in the months ahead: if critical bugs keep surfacing and getting fixed promptly, the model is holding; a conspicuous quiet spell would raise the opposite worry. Third, how the VIP tier is populated — whether it stays genuinely meritocratic and accessible to newcomers who prove themselves, or hardens into a closed circle. Bug bounties work because the pool of people looking is large and motivated. The bet GitHub just placed is that a smaller, better-paid pool finds more of what matters. Whether that bet pays off is the question the rest of the industry will be watching.