1、Copyright 2025 S&P Global.All Rights Reserved.Lessons from industry leadersSecuring AI at scaleMay 2025451 Research Discovery ReportSecuring AI at scale| of contentsIntroduction3Methodology3Figure 1:Demographic breakdown of participants3Insight 1:Security for AI should span multiple environments and

2、 be designed to scale4Figure 2:Estimated percentage of AI workloads by execution platform5Insight 2:Additional security tools and capabilities are sorely needed7Figure 3:Anticipated infrastructure changes to support AI/ML use cases over the next 12-24 months7Insight 3:Security tools and strategies s

3、hould support diverse governance requirements9Insight 4:AI models themselves must be secured and monitored,in addition to the environments where they run11Figure 4:Organizational AI maturity12Implications and next steps13Insight 1:Security for AI should span multiple environments and be designed to

4、scale13Insight 2:Additional security tools and capabilities are sorely needed13Insight 3:Security tools and strategies should support diverse governance requirements14Insight 4:AI models themselves must be secured and monitored,in addition to the environments where they run14About the author15Securi

5、ng AI at scale| AI adoption continues at a fever pitch,organizations are experiencing a variety of headwinds that prevent them from realizing its full value.According to 451 Researchs Voice of the Enterprise:AI&Machine Learning,Infrastructure 2024 survey,security issues present a major technological

6、 barrier inhibiting organizations AI efforts.Improvements are needed,not only in the tools enterprises use to help secure their environments,but also in organization-wide awareness and training to develop core competencies for AI security.This report is the second in a three-part series about AI at