HackerOne's move towards agentic triage raises some interesting questions about the future of bug bounty programs.
AI can potentially make triage faster and reduce backlogs, but security reports aren't always straightforward. Two reports can look similar on the surface while having completely different root causes, exploitability or impact.
I'm curious how the community feels about this:
Are AI-driven duplicate decisions reliable enough to close reports automatically?
Can these agents properly understand complex vulnerabilities and technical evidence?
Is faster triage worth the possibility of incorrect decisions?
Should researchers always have access to human review when disputing an automated closure?
How transparent should platforms be about why an AI agent made a particular decision?
Who should ultimately be accountable when an automated triage decision is wrong?
I'm not against AI being part of the process.…