HackerOne Halts New Report Submissions Amid Flood of AI-Generated Vulnerability Reports — Bug Bounty System Shaken
HackerOne, the US security company that has offered bounties to encourage the discovery of vulnerabilities in open source software (OSS), has stopped accepting new reports. According to ITmedia, the cause is a surge in low-quality vulnerability reports generated by AI.
With the spread of AI tools, even users without security expertise can now automatically scan for and report vulnerabilities. However, most of these reports cover known issues or false positives, reportedly consuming significant security team resources. The impact extends to major OSS projects, and Google, which runs a similar bounty program, is also being forced to respond.
Bug bounties have long been a key pillar of the security ecosystem. But the flood of AI-generated noise reports has begun to shake that foundation, making quality control methods for reports an urgent priority for the AI era.
PS3 Emulator Developers: "Please Stop Sending AI-Generated PRs" — Open Source's New Headache
The development team behind a PlayStation 3 emulator has raised complaints about a flood of pull requests (PRs) generated with AI coding tools. According to Kotaku, the story has struck a chord with the community, picking up 48 points on Hacker News.
With the spread of AI coding assistants, even programming beginners can now easily create PRs. But AI-generated PRs that show no understanding of the code's intent increase review burden and exhaust project maintainers. For complex systems like a PS3 emulator, changes submitted without a deep understanding of the architecture can end up being counterproductive.
The problem is not limited to this project. As AI coding tool usage grows, similar challenges are expected to surface across many OSS projects. Just as with the HackerOne situation, this exposes a structural issue: the AI-driven "explosion of volume" is straining existing quality control processes.
Microsoft Releases Agent Governance Toolkit — Putting a "Gatekeeper" on AI Agents
Microsoft's Agent Governance Toolkit (AGT), released as open source in April 2026, is also drawing attention in Japan's developer community. According to an article on Zenn, AGT serves as a "gatekeeper" that controls who can do what, and under which conditions, when AI agents autonomously invoke tools.
As AI agents gain broader permissions — reading and writing files, calling APIs, querying databases — governance mechanisms are essential for deploying them in production. AGT provides a .NET (C#) implementation, allowing agent behavior to be controlled on a policy basis.
How to balance AI agent autonomy with safety is one of the key technical challenges of 2026. The arrival of an official toolkit like AGT can be seen as a step toward standardizing agent operations.
The AI Customer Support Paradox — Cost Cutting That Destroys Customer Experience
Another challenge in operating AI agents has also been pointed out. An article published on Zenn analyzes that "when you hand initiatives over to AI agents, sales and click-through rates tend to grow, but unsubscriptions, churn, and complaints can also increase," highlighting the risk that optimizing short-term metrics damages medium- to long-term customer relationships and emphasizing the importance of "guardrail metrics."
On Hacker News, a discussion titled "AI Productivity Fails" reflects growing interest in the problems playing out behind AI-driven productivity gains. James Shore has also argued that "what we need is AI that reduces maintenance costs," prompting a rethink of the direction of AI investment.
---
Flood of AI-Generated "Garbage Reports" Leaves Teams Overwhelmed and Exhausted... Something Is Changing in Vulnerability Bounty Programs (ITmedia AI+)
PS3 Emulator Devs Politely Ask That People Stop Flooding It with AI PRs (Kotaku / Hacker News)
Putting a "Gatekeeper" on AI Agents — Introduction to the Agent Governance Toolkit (.NET Edition) (Zenn)
AI Agents Can Grow Sales While Destroying Customer Experience (Zenn)
AI Productivity Fails (Hacker News)
You Need AI That Reduces Maintenance Costs (James Shore / Hacker News)