Collaborative Industry Effort Delivers Identity Management, Supply Chain Integrity, and Protocol Security for AI Agent Deployments
OASIS Open, the international open source and standards consortium, announced the release of the “Model Context Protocol (MCP) Security” white paper from the Coalition for Secure AI (CoSAI), an OASIS Open Project. This framework equips security professionals and developers to identify, assess, and mitigate risks in MCP-based AI agents, addressing the urgent need for standardized security practices as AI increasingly connects to external tools and services.
Securing the Bridge Between AI and the Real World
MCP, developed by Anthropic, a CoSAI Sponsor, together with a growing open source community, has emerged as a leading protocol for connecting AI agents to external tools, databases, APIs, and services. However, like any integration protocol, MCP deployments face active and evolving threats.
This security framework presents a well-defined taxonomy of nearly forty threats and concrete mitigation strategies across twelve distinct categories, spanning identity and access control, input validation, data protection, network security, supply chain integrity, and operational visibility. The framework distinguishes between traditional security concerns amplified by AI mediation and novel attack vectors unique to LLM-tool interactions, enabling security teams to implement defense-in-depth strategies tailored to their specific deployment patterns.
Also Read: CIO Influence Interview with Gera Dorfman, Chief Product Officer at Orca
“As AI moves beyond chat models to agents, gaining the ability to take actions and interact with their environments and the real world through tool calling, the security implications and potential consequences are much more severe,” said Ian Molloy, IBM, and Sarah Novotny, CoSAI’s Workstream 4 Co-Leads. “This framework represents the expertise of CoSAI’s members and contributors who understand that protecting agentic systems requires addressing everything from protocol-level authentication and supply chain integrity to guardrails, systems security and enforcement.”
Collaborative Industry Effort
The MCP Security paper was developed by CoSAI’s Workstream 4: Secure Design Patterns for Agentic Systems, drawing on contributions across CoSAI’s Sponsors and partner organizations, including Premier Sponsors EY, Google, IBM, Meta, Microsoft, NVIDIA, PayPal, Snyk, Trend Micro, and Zscaler. Additional CoSAI AI Security Guidance Publications can be found on GitHub (https://github.com/cosai-oasis/).
Technical contributors, researchers, and organizations are welcome to participate in CoSAI’s open source community and support its ongoing work. OASIS welcomes additional sponsorship support from companies involved in AI security. Contact join@oasis-open.org for more information.
Catch more CIO Insights: Identity is the New Perimeter: The Rise of ITDR
[To share your insights with us, please write to psen@itechseries.com ]
