Leading cybersecurity teams are increasingly using artificial intelligence agents to support human operators rather than replace them, according to new research from cyber-readiness platform Hack The Box.
The platform’s three-year benchmark data found that AI agents are disproportionately used by some of the strongest-performing teams in its cybersecurity challenges.
Among the top 25 teams, 68% had at least one AI agent, despite AI agents accounting for only 2.7% of all registered accounts.
These agents were responsible for 4.2% of submitted flags and 4.6% of points awarded.
AI-assisted teams solve challenges faster
Across all active teams included in the research, AI-augmented teams recorded a 3.2-times higher solve rate than teams without AI assistance.
The advantage narrowed to 1.69 times among the top 5% of teams, but their speed advantage increased.
AI-augmented teams completed challenges three to four times faster, according to Hack The Box.
However, the data also showed that AI did not necessarily outperform the strongest human teams in overall completion.
The strongest human team completed all 36 challenges, while the strongest AI-assisted team completed 32.
Human expertise remains critical
Researchers said AI delivers the greatest value when it is guided by people who already understand the underlying cybersecurity problem.
AI agents can help accelerate research, coding, troubleshooting and initial analysis, but human operators remain responsible for selecting strategies, questioning weak outputs and determining whether results are accurate.
Hack The Box cautioned that the findings do not prove that AI itself caused teams to perform better.
Instead, the data indicates that AI is already becoming part of the toolkit used by many of the strongest cybersecurity practitioners.
Cybersecurity teams solving challenges faster
The research also found a significant improvement in overall performance during the past three years.
The median time taken to solve challenges fell from 26.1 hours in 2024 to 13.8 hours in 2026.
At the same time, more teams managed to complete the entire challenge board despite the number of challenges increasing.
Only two teams completed the full board in 2024, followed by three in 2025. In 2026, that number increased to 15 teams.
Hack The Box said the improvement could reflect better preparation, reusable tools, greater familiarity with the platform and deeper specialist expertise.
AI is becoming part of security operations
Haris Pylarinos, founder and CEO of Hack The Box, said the debate is increasingly shifting from whether AI will become part of cybersecurity operations to whether security teams can use it safely and effectively.
He said the data showed AI appearing most often alongside highly skilled practitioners rather than replacing them.
As AI agents become more capable, Pylarinos argued, human judgement, validation and technical expertise will become increasingly important.
