SEC536: Adversarial AI - Penetration Testing AI Systems


Experience SANS training through course previews.
Learn MoreLet us help.
Contact usBecome a member for instant access to our free resources.
Sign UpWe're here to help.
Contact UsKey findings:
The findings point to a persistent gap between automation ambition and operational reality: organizations are investing heavily in tools and planning to expand AI and ML use, yet nearly all still depend on human intervention somewhere in the response chain. Budget constraints and a shortage of skilled personnel are the two forces most often cited as slowing that transition, and the near-even split on team structure suggests the industry has not converged on a single best model for organizing detection and response work. Respondents came primarily from the cybersecurity, banking and finance, government, and education sectors, spanning organizations from fewer than 1,000 employees to more than 50,000, with roles ranging from incident responders and SOC analysts to security managers and directors.
Only 16% of organizations report having fully automated their response processes, while 64% have partially integrated automated response mechanisms.
The need for skilled personnel is the top obstacle, cited by 59% of respondents, followed closely by budget constraints at 47%.
Only 22% of respondents rate AI/ML-based detection tools as extremely effective, while 21% find them ineffective, placing AI/ML roughly in the middle compared to other detection technologies.
Organizations are nearly evenly split: 48% use separate specialized teams and 48% operate as a single integrated team, indicating no industry consensus on the best structure.
41% of organizations can respond to confirmed threats within minutes, and 8% within seconds, though 12% take a day or longer.


Josh leads global MDR at Uptycs, defending major international brands, while also serving as an independent DFIR expert advising legal, government, and commercial clients in Australia.
Read more about Josh Lemon


















