Security

AI Agent RCE Flaws in Semantic Kernel Explained

3 min read

Summary

Microsoft Defender researchers disclosed two fixed vulnerabilities in Semantic Kernel that could let prompt injection escalate into host-level remote code execution in AI agents. The findings matter because they show how unsafe tool parameter handling in agent frameworks can turn natural language inputs into code execution paths, raising the stakes for organizations building or securing AI-powered apps.

Need help with Security?Talk to an Expert

AI Agent RCE Flaws in Semantic Kernel Explained

Introduction

AI agents are changing enterprise application design, but they also introduce a new execution risk. Microsoft Defender Security Research has detailed how prompt injection in AI agent frameworks can move beyond content manipulation and become host-level remote code execution (RCE) when tools and plugins trust model-generated parameters.

For security teams and developers using agent frameworks, this is an important reminder: once an LLM can call tools, weaknesses in framework logic can directly affect the underlying system.

What’s new

Microsoft disclosed two critical vulnerabilities in the open-source Semantic Kernel framework:

  • CVE-2026-26030: An RCE path involving the In-Memory Vector Store when used with the Search Plugin in its default configuration
  • CVE-2026-25592: An arbitrary file write issue through SessionsPythonPlugin

According to Microsoft, both vulnerabilities have been fixed.

The most notable finding is that exploitation did not require a browser exploit, malicious attachment, or memory corruption bug. In the demonstrated scenario, a single prompt injection was enough to influence tool parameters and trigger code execution on the host.

Why the issue occurred

The research highlights a broader design problem in AI agent frameworks:

  • Agents interpret natural language and map it to tool calls
  • Frameworks often trust parsed model output too much
  • Unsafe parameter handling can create execution sinks
  • Blocklist-based protections can be bypassed in dynamic languages like Python

In the Semantic Kernel case, Microsoft researchers found unsafe string interpolation in a Python lambda expression executed with eval(), combined with a validator that could be bypassed.

Impact on IT and security teams

Organizations experimenting with AI agents, copilots, or custom LLM apps should treat this as a framework security issue, not just an AI safety issue.

Potential exposure is highest where:

  • Semantic Kernel is used in production or internal apps
  • Agents can access plugins, scripts, files, or data stores
  • Prompt injection is possible through user input, documents, or connected content sources
  • Default Search Plugin and In-Memory Vector Store configurations are in use

This research also has implications beyond Semantic Kernel. Many teams use frameworks such as LangChain, CrewAI, or similar orchestration layers, and the same trust model concerns may apply.

Security and platform teams should:

  • Patch affected Semantic Kernel deployments immediately
  • Inventory AI agents and plugins that can execute code, read files, or access sensitive systems
  • Review tool-calling paths for unsafe deserialization, interpolation, or dynamic execution patterns
  • Harden prompt injection defenses and assume hostile input can reach agent tools
  • Audit logs and telemetry for suspicious plugin invocations or unexpected process execution
  • Reduce agent privileges so successful prompt injection cannot easily lead to system compromise

Bottom line

Microsoft’s research shows that in agentic applications, prompt injection can become an execution primitive when frameworks and tools over-trust model output. For defenders, the priority is clear: patch vulnerable frameworks, review plugin design, and apply least privilege before AI agents become a new RCE surface.

Need help with Security?

Our experts can help you implement and optimize your Microsoft solutions.

Talk to an Expert

Stay updated on Microsoft technologies

Read the original article by Microsoft Defender Security Research Team, Uri Oren, Amit Eliahu and Dor Edry
Semantic KernelAI agentsprompt injectionremote code executionMicrosoft Defender

Related Posts

Security

AI Memory Security in Microsoft 365 Explained

Microsoft has outlined how it secures AI memory in Microsoft 365, addressing emerging risks such as memory poisoning and delayed tool execution. The update matters because persistent AI memory can improve personalization and agent performance, but it also creates new security, compliance, and audit requirements for IT and security teams.

Security

Parallel Threat Activity: Microsoft DART Findings

Microsoft Incident Response detailed a complex intrusion in which two unrelated threat actors operated simultaneously in the same environment, complicating attribution and detection. The case highlights how ransomware activity, SharePoint exploitation, trusted tool abuse, and identity compromise can overlap across hybrid estates, reinforcing the need for strong telemetry, patching, and coordinated response.

Security

AutoJack RCE in AutoGen Studio: Security Lessons

Microsoft security researchers detailed AutoJack, an exploit chain in AutoGen Studio that could let untrusted web content rendered by an AI browsing agent trigger remote code execution on the host. Although the vulnerable MCP WebSocket surface was never shipped in a PyPI release and the issue was hardened upstream during development, the findings highlight important security risks for agent frameworks that combine web browsing with privileged local services.

Security

Microsoft Security Forrester Study Reports 124% ROI

A new Forrester Total Economic Impact study found that organizations consolidating on Microsoft Security could see a projected 124% ROI over three years. The report highlights lower breach risk, reduced remediation costs, lower technology spend, and productivity gains as key reasons unified security platforms matter in the AI era.

Security

Mastra npm Supply Chain Attack: What IT Teams Need to Know

Microsoft has detailed a large-scale npm supply chain compromise affecting more than 140 Mastra packages after an attacker took over a maintainer account and injected a malicious dependency. The attack is significant because the payload executed during npm install, putting developer workstations and CI/CD pipelines at risk even if the package was never directly used in code.

Security

Crypto Clipper Malware Uses Tor and USB Worm Spread

Microsoft has detailed a Windows-based crypto clipper campaign that uses malicious shortcut files, a bundled Tor client, and worm-like USB propagation to steal wallet data and maintain persistence. The threat matters because it combines clipboard theft, screenshot exfiltration, and remote code execution with stealthy Tor-based command and control, making behavioral detection critical for defenders.