How the AI Era Is Fueling a Bug Hunting Arms Race

A decade ago, the concept of paying researchers for uncovering software flaws was just beginning to gain traction. Vulnerability disclosure and bug bounty programs marked a major shift, moving organizations away from hostility toward security researchers and toward a more cooperative model of receiving input and releasing fixes. When Apple finally launched its bug bounty in 2016, the top reward was $200,000. That figure jumped to $1 million by 2019 and reached $2 million last year. Now, the entire ecosystem is poised for another transformation.

As agentic AI models grow more capable of both autonomously identifying software vulnerabilities and developing exploits,essentially finding weaknesses and building hacking tools,vulnerability disclosure programs are being inundated. Organizations are discovering more bugs than ever, but this abundance is reshaping the economics of bug bounties for everyone involved. That includes institutions soliciting submissions and researchers, many of whom rely on bug hunting for income or supplemental earnings. Crucially, the same changes are unfolding for attackers.

“I’ve probably submitted three times more bugs than I did last year at this time,” says independent security researcher Joseph Thacker, who has developed AI-driven methods and tools for his own bug hunting. “I would suspect that a company like Google is going to spend two to 10 times as much on bug payouts as they did last year.”

Tech giants, he adds, “can handle that pressure, but most companies can’t. Right now people will be submitting low- and medium-hanging fruit,agents are finding really good bugs. But next year there will be fewer bugs submitted because a lot of that will already have been found, and I think some companies will up their payouts again.”

Thacker and other researchers admit that no one knows exactly how the supply and demand dynamics will play out long term. Depending on how effective AI-driven exploit discovery and automated scanning become for attackers, developers may feel even more pressure to release patches quickly. That could accelerate longstanding standards like the 90-day disclosure deadline, a window between finding a bug and disclosing it publicly that often spurs patch releases.

Security researcher Himanshu Anand wrote earlier this month, “The 90 day responsible disclosure window was built for a world where bug finders were rare and exploit development was slow. That world is gone. LLMs have compressed both timelines.”

Crucially, forced accountability from attackers could also motivate organizations to speed up how they deploy vulnerability fixes. Patch management has always been a complex security challenge, as installing new software at scale without proper testing can lead to unintended consequences, including worst-case scenarios like outages.

The urgency of real-world attacks enabled by AI appears to be growing. Both sophisticated and less-proficient actors are looking to expand their capabilities and cut costs. In findings published earlier this month, for example, Google researchers reported observing “prominent cyber crime threat actors” (whom they declined to name) attempting to exploit a zero-day vulnerability that they had developed using AI tools. The goal was to bypass two-factor authentication on an open source system administration platform. Google notified the developer quickly, and a fix was issued. But the researchers said the incident was a crucial illustration of the changing bug-hunting landscape.

“We all assumed it was already happening, and this is our first evidence that it is happening,” says John Hultquist, chief analyst for Google Threat Intelligence Group, referring to attackers using AI to discover novel vulnerabilities and create exploits.