AI-Generated Bug Reports Overwhelm Bug Bounty Programs

• •Bug bounty programs are inundated with low-quality AI-generated reports, forcing suspensions.
• •GitHub has tightened its submission standards due to the rise in speculative AI-assisted reports.
• •The AI model Mythos is contributing to the flood of unvalidated vulnerability submissions.

Bug bounty programs are facing a significant surge in low-quality AI-generated vulnerability reports, leading some companies to suspend their initiatives. Reports submitted through platforms like Bugcrowd have quadrupled in volume, with a majority deemed false or misleading. GitHub and other industry leaders are tightening submission standards, requiring proof-of-concept and validation to combat the influx of speculative reports. Companies such as Curl and Nextcloud have temporarily halted their bounty programs due to the overwhelming number of unvalidated submissions. The AI model Mythos from Anthropic is at the center of this issue, as it can quickly generate numerous vulnerability reports. Security experts warn that while AI tools can assist in identifying vulnerabilities, they also increase the burden on human researchers to filter out noise from genuine threats. The situation highlights a growing tension in the cybersecurity landscape as AI tools become more prevalent.