MCP is an application-level protocol that provides a simple and agreed upon messaging pattern and transport format currently used by many AI-enabled systems for managing interactions between services. The guidance aims to reduce risk while supporting safe innovation in AI-augmented systems.

Real-world adoption of MCP has accelerated. It is increasingly found in AI deployments across products used in business, finance, legal, software development, and other industries, including for sensitive tasks like querying personally identifiable information.

While MCP simplifies the integration of diverse capabilities into powerful agent workflows, the current protocol specification requires careful and cautious implementation for security. Gaps in MCP design, implementation, and operational posture have created significant and evolving security concerns including serialization risks, trust boundaries, and agent misuse, to name a few, according to the CSI.

Although traditional cybersecurity principles such as authentication, authorization, and input validation remain necessary protective measures, agentic AI systems — especially those featuring MCP — introduce novel and systemic risks like dynamic tool invocation, implicit trust relationships, and context sharing. Established cyber defense strategies unfortunately do not adequately address these new risks.
These are not isolated problems that can be patched at the interface or endpoint level. Securing MCP systems requires treating the agentic environment as a continuum. Misaligned assumptions or subtle inconsistencies at any stage can propagate and compound into exploitable conditions.

This report examines these security concerns, outlines gaps that must be addressed before MCP can be used securely and confidently. It offers practical recommendations for organizations adopting MCP in high-stakes or production environments. The guidance is designed to remain relevant as the MCP protocol, implementations, and operations continue to evolve.

Adopters are advised to proceed with caution, drawing on lessons from prior distributed and plugin-based ecosystems while applying heightened scrutiny to MCP’s novel integration and automation patterns. Continued collaborative work among implementers, security researchers, and standards organizations will be essential to establish more robust and trustworthy foundations for AI infrastructure, particularly for national security and other high assurance environments.

Read the full report.

Source: NSA

Stay in the know with breaking news from across the IC and IC contracting landscape by becoming a paid subscriber to IC News. Your support makes our work possible.

The post NSA releases security design considerations for AI-driven automation appeared first on Intelligence Community News.

The agencies are releasing the report to help developers, providers, and systems owners develop, deploy, and operate secure Artificial Intelligence (AI) systems, including those used in National Security Systems (NSS), by the Department of Defense (DoD), and by the Defense Industrial Base (DIB).

“We wish we could rewind time and bake security into the start of the internet. We have that opportunity today with AI. We need to seize the chance,” said Rob Joyce, NSA cybersecurity director.

According to the CSI, AI systems are subject to security vulnerabilities that need to be considered alongside standard cyber threats. For example, AI systems are vulnerable to “adversarial machine learning” (AML) attacks, which exploit fundamental vulnerabilities in machine learning (ML) systems, including hardware, software, workflows, and supply chains. Prompt injection and training data poisoning are examples of AML attacks that could enable malicious cyber actors to compromise an ML model’s classification or regression performance, perform unauthorized actions, or extract sensitive information.

The CSI indicates that secure by design principles are applicable to AI systems. Providers of AI components should implement security controls by design and default within their ML models, pipelines, and systems. Accordingly, the CSI focuses on four key areas of AI system development: secure design, secure development, secure deployment, and secure operation.

The UK National Cyber Security Centre (NCSC-UK) and the U.S. Cybersecurity and Infrastructure Security Agency (CISA) co-authored the CSI with NSA and other partners.

The authoring agencies advise that this CSI does not replace general cybersecurity best practices and risk management programs. Recommendations in the CSI should be considered in conjunction with established cybersecurity, risk management, and incident response best practices.

Read the full report.

Source: NSA

Your competitors read IC News each day. Shouldn’t you? Learn more about our subscription options, and keep up with every move in the IC contracting space.

The post NSA and partners issue guidance for securing AI appeared first on Intelligence Community News.