Industry Leadership
Strategic Initiatives
CSA's strategic programs driving innovation in AI, cloud, and Zero Trust.
A public-interest 501(c)(3) dedicated to secure and trustworthy AI.




Industry Leadership
Strategic Initiatives
CSA's strategic programs driving innovation in AI, cloud, and Zero Trust.
A public-interest 501(c)(3) dedicated to secure and trustworthy AI.

CSAI FoundationChaptersEventsBlog

CSA Research

Best practices, guidance, frameworks and tools to help the industry secure the cloud. Read our research to get your questions around cloud security answered.
Research

CSA Research is created by the industry for the industry and is both vendor-neutral and consensus driven. Our research is created by subject matter experts who volunteer for our working groups. Each working group focuses on a unique topic or aspect of cloud security, from IoT, DevSecOps, Serverless and more, we have working groups for over 20 areas of cloud computing. You can view a list of all active research working groups. To find out more about how our research is created and the process we follow you can view the CSA Research Lifecycle.

Contribute to CSA Research

Peer reviews allow security professionals from around the world to collaborate on CSA research. Provide your feedback on the following documents in progress.

Latest Research

Top Threats to Cloud Computing Survey Report 2026

Top Threats to Cloud Computing Survey Report 2026

Release Date: 08/12/2026

CSA's 2026 Top Threats report reveals a decisive change in cloud security priorities over the last two years. Based on a global survey of industry professionals, the report identifies today's 11 most critical cloud security issues. Identity, AI, third-party resources, and API threats dominate...
AISI: Open-Weight Models Close Cyber Capability Gap

AISI: Open-Weight Models Close Cyber Capability Gap

Release Date: 08/09/2026

Key Takeaways The UK AI Security Institute (AISI) has found that the leading open-weight AI models now trail frontier closed models on offensive cyber tasks by only four to seven months, down from a six-to-ten-month lag measured through most of 2025 [1][2].
Four AI Escapes: A Systemic Governance Risk Reading

Four AI Escapes: A Systemic Governance Risk Reading

Release Date: 08/09/2026

What Sandbox and Containment Failures at OpenAI and Anthropic Reveal About Evaluation Governance. Key Takeaways Between July 21 and July 30, 2026, OpenAI and Anthropic separately disclosed four distinct incidents in which frontier models breached the containment boundaries of their own...