Around 45% say that AI is moderately beneficial and they’re starting to note the benefits

Looking ahead, 70% of organisations will focus on AI/ML data usage governance.

False positive and negative rates are the No. 1 way that organizations reported that they evaluate the efficacy of AI in security, named by 66% of respondents

Almost 60% of organisations added new AI data responsibilities in the past year.

83% of CISOs place significantly higher priority on AI data usage governance.

86% of security teams today utilize some type of AI within their security tool stack

55% of Security Managers/Directors added data governance duties for AI training.

52% of Security Engineers/Architects have new AI data discovery responsibilities.

Security and privacy risks were a reason for turning off AI functionality, cited by 55%

46% of firms say that they’re actively trying to use AI to solve false positive issues

Approximately 56% of respondents reported that at least half of their security vendors tout their AI capabilities

Basic vulnerability scanning was a current application of AI in 47% of security tech stacks

A lack of transparency and explainability was the top reason for turning off AI functionality, cited by 58%

100% of those who said their AI is very beneficial and a vital part of their security program have internal data science staff members

Nearly a third of respondents reported that their team spends at least four hours per week training AI models within their own tools or within commercially available AI functionality

83% of Security Engineers/Architects worry most about AI systems understanding data access rights.

77% of security teams cannot ensure AI systems respect proper data access rights.

72% of CISOs express the most concern about discovering data used in AI initiatives.

Antivirus/anti-malware was a current application of AI in 40% of security tech stacks

The top five vulnerability management problems they’re actively trying to solve with AI today were: false positives (49%), overload of data (39%), reliance on manual processes (33%), disparate results from scanning tools (31%), and false negatives (31%)