Patchstack Radically Restricts Bug Bounty Program to Combat Flood of AI-Generated Vulnerability Reports

Patchstack, a company that runs a significant bug bounty program for the WordPress ecosystem, has announced sweeping changes to its vulnerability submission rules in response to a massive increase in low-impact, AI-generated reports. From June 1, 2026, submissions will only be accepted if they meet one of four narrow sets of criteria, aiming to refocus the program on vulnerabilities that pose genuine, widespread risk to websites. The move underscores how artificial intelligence is reshaping security research, forcing bug bounty operators to adapt to an era where automation can rapidly produce superficially valid but practically shallow findings.

The shift is not sudden. Historically, Patchstack, which is also a CVE Numbering Authority (CNA) — an organization authorized to assign official identifiers to publicly disclosed security flaws — kept its bounty scope aimed at clear, meaningful threats. But throughout 2026, the volume and nature of reports changed dramatically. AI tools have enabled researchers to churn out submissions that just barely fit the program’s formal requirements while lacking real-world impact. Many reports contain incorrect AI assumptions, such as misclassifying a limited file upload as a full arbitrary upload, or assuming a nonce — a secret token used to prevent unauthorized actions — can be guessed remotely. The result is a flood of data that bogs down triage teams.

Patchstack’s internal statistics paint a vivid picture. A graph of report outcomes over recent months shows sharp spikes in rejected and duplicate submissions. For some vulnerabilities, over 20 duplicate reports landed from different researchers, demonstrating the speed at which AI can replicate known vulnerability patterns at scale. While AI has also improved the quality of some proof-of-concepts and helped uncover more flaws overall, the program’s sustainability was threatened. This mirrors a wider trend: many bug bounty programs have been paused or restructured as the industry grapples with AI’s dual-edged impact.

The new rules, effective June 1, 2026, aim to filter out low-value noise. To qualify, a report must now meet at least one of four criteria. First, it can be a true zero-day, meeting the program’s existing zero-day requirements. Second, it must involve authentication at the level of guest, subscriber, customer, or similar roles not explicitly granted by an admin; the contributor role is no longer in scope. Third, the vulnerability must be of a specific type and satisfy strict conditions — for example, Cross-Site Scripting (XSS) must affect the entire site or be reflected; arbitrary file uploads require full control over both path and extension; privilege escalation must reach at least contributor access; and IDOR (insecure direct object reference) must have significant security impact, not just leak personal information. Fourth, reports on software within Patchstack’s managed Vulnerability Disclosure Program (mVDP) scope will still be accepted, including those involving contributor roles, provided they demonstrate a measurable security impact.

Alongside the submission criteria, Patchstack is overhauling its reward structure. Bounties will no longer be issued based on predefined levels, and random prizes for selected researchers are eliminated. The monthly bounty pool is now limited to the top five submissions. Reports affecting mVDP software may still be accepted but might not earn experience points. Additionally, researchers who submit reports that clearly violate the rules, contain unverified AI-generated assumptions, or were not tested against the actual software will receive an immediate one-week ban from the program.

For website owners and IT teams, these changes highlight an essential truth: the sheer volume of vulnerabilities being discovered means proactive defense is critical. While bug bounty programs help surface flaws, the onus is on site operators to ensure their platforms are patched and hardened. Using a managed WordPress hosting service like AEU Hosting, which includes automatic updates, malware scanning, and firewall protection, can significantly reduce exposure to the kinds of threats that bug bounty programs are designed to identify before attackers exploit them.