Most of the top security teams are using AI agents, employing them to support human activity rather than replace it, according to Hack The Box.
The latest three-year benchmark data from the cyber readiness platform finds disproportionate use of agents among many of the strongest-performing teams.
More than two-thirds (68%) of the top 25 teams having an AI agent, although AI agents represented just 2.7% of registered accounts.
Those agents accounted for 4.2% of submitted flags and 4.6% of points awarded.
Across all active teams in the research, AI-augmented teams recorded a 3.2 times solve-rate advantage; an advantage that narrowed to 1.69 times among the top 5%, while the speed advantage grew.
While AI-augmented teams completed challenges three to four times faster, with the speed advantage increasing among the strongest operators, the strongest human team completed all 36 challenges compared with 32 for the strongest AI team.
"AI therefore creates the greatest value when it is directed by people who already understand the problem. It can compress research, coding, troubleshooting, and first- pass analysis. The operator still has to choose the path, challenge weak output, and decide whether the result is valid," the researchers said.
AI is now a critical component in security operations
Together, the findings suggest that AI is no longer just a question of experimentation, but is becoming part of the working methods of experienced cybersecurity practitioners.
Researchers emphasized the importance of human expertise in directing, evaluating, and validating AI-generated work, however.
"The findings do not establish that AI caused teams to perform better. They do, however, show that AI is already becoming part of the toolkit used by many of the competition’s strongest teams," Hack The Box said.
Performance across the competition has shifted substantially over the past three years, with the median recorded time-to-solve dropping from 26.1 hours in 2024 to just 13.8 hours this year.
At the same time, and even with the challenge board expanding, the number of teams completing the entire challenge board rose from two in 2024 and three in 2025 to 15 in 2026.
The improvement, said Hack The Box, could be down to better preparation, reusable tooling, platform familiarity, and deeper specialist knowledge.
“The question for security leaders is no longer whether AI will become part of cybersecurity operations. That is already happening on both sides of the equation,” said Haris Pylarinos, Founder and CEO of Hack The Box.
“What matters now is whether teams have the expertise to use it safely and effectively. Our data shows that AI is appearing most often alongside some of the strongest practitioners, not instead of them. As agents become more capable, human judgment, validation and hands-on technical skill become more important, not less.”
FOLLOW US ON SOCIAL MEDIA
Follow ITPro on Google News and add us as a preferred source to keep tabs on all our latest news, analysis, views, and reviews.
You can also follow ITPro on LinkedIn, X, Facebook, and BlueSky.
Nasuni expands unstructured data platform with DryvIQ acquisition
Three things you need to know about Claudeforce
