Skip to main content
Mallory
← Back to News
ai-platform-securitycommand-and-control-methoddata-exfiltration-methodinitial-access-method

MCP Tool Attacks Expose Autonomous Agents to Secret Theft and Unauthorized Actions

Emerging Technology Risk · 10 sources · Updated 2mo ago · First seen Sep 19, 2025

MCP Tool Attacks Expose Autonomous Agents to Secret Theft and Unauthorized Actions

Model Context Protocol (MCP) tools are expanding the attack surface of LLM-powered autonomous agents by connecting them to external functions, services, and sensitive data. Attackers can exploit conventional implementation defects—including reported command-injection and CSRF issues—or manipulate agent behavior through indirect prompt injection hidden in tool metadata, parameters, retrieved content, and multi-tool workflows. Documented techniques include tool poisoning, encoded or invisible instructions, tool redefinition “rug pulls,” tool-name collisions, passive influence, and orchestration injection.

Successful MCP attacks can cause secret exfiltration, unauthorized command execution, financial fraud, and publication of private data. Organizations deploying MCP-enabled agents should restrict tools to trusted sources, enforce least-privilege access and sandboxing, require human approval for sensitive operations, disable auto-run and persistent “always allow” settings, manually review tool definitions and updates, and maintain auditable clients with comprehensive tool-invocation logging.

New surface, new adversariesTrack how attackers are adapting to this technology.
Track this risk
Event timeline

How this story unfolded

Events from the most recent confirmed update back to the earliest known activity.

7 events · Mar 2025 – Sep 13, 2026
  1. Reported Sep 16, 20262mo ago

    MITRE ATLAS documents Jira ticket prompt injection via MCP tools

    MITRE ATLAS case AML.CS0039 describes an attacker embedding prompt-injection instructions in a public Jira Service Management ticket. An AI support agent processing the ticket through authorized Atlassian MCP tools was induced to collect accessible Jira data and return it in a response to the attacker-controlled ticket, without a CVE or conventional software exploit.

    Безопасность AI агентов: атаки через MCP и детект
  2. Sep 13, 20262mo ago

    Paper formalizes MCP stochastic-deputy tenant-isolation risk

    Researchers submitted a paper describing how attacker-controlled context can induce tool-using LLM agents to select resources outside a caller's intended tenant scope when MCP tools accept tenant identifiers. Across 373 trials, validated tenant parameters still allowed all 26 observed out-of-scope attempts in one subset; the paper proposes removing tenant identity from tool schemas and enforcing credential-bound scope below the agent layer.

    [2609.14780] The Stochastic Deputy: Structural Tenant Isolation for Tool-Using LLM Agents
  3. Reported Sep 11, 20262mo ago

    Pentest demonstrates MCP tool-description prompt injection stealing SSH key

    A pentest described a hidden instruction in an MCP tool's description that caused an agent to read ~/.ssh/id_rsa and return the private key in a chat session, without exploiting a conventional CVE. The demonstration highlights MCP tool metadata as a prompt-injection and credential-exfiltration vector.

    Атаки на MCP AI-агентов: 31 метод и PoC
  4. Jul 28, 20264mo ago

    MCP specification adds stronger authorization controls

    MCP maintainers released a specification update emphasizing issuer validation, issuer-bound client credentials, and Client ID Metadata Documents as the preferred client-registration method, strengthening the protocol's authorization model.

    Researchers found that 1 in 5 MCP access policies came back broken or missing - The New Stack
  5. Jul 2026

    Island identifies instruction-supply-chain risks in MCP server builds

    In July 2026, Island analyzed 33,563 MCP server builds and found apparent instruction or output-manipulation signals in about 3.3% of them, while assessing 49% as having potential manipulation capability. The analysis also highlighted covert logging instructions in a marketing-analytics MCP dependency and a postmark-mcp update that added a BCC backdoor copying outbound email to an attacker-controlled address.

    When English becomes exploit code: The hidden risk inside MCP servers | resource | SC Media
  6. May 2025

    Prompt injection against GitHub MCP server exposes private repositories

    An attacker reportedly used prompt injection against a GitHub MCP server to obtain private repository data. The incident was attributed to a personal access token whose permissions exceeded those needed for the requested task.

    Why MCP security is about permissions overhaul - The New Stack
  7. Mar 2025

    Researchers find widespread flaws in public MCP server implementations

    Researchers examining publicly available MCP server implementations reported command-injection flaws in 43% of tested implementations and unrestricted URL fetching in 30%.

    MCP Tools: Attack Vectors and Defense Recommendations for Autonomous Agents
Do you run this?Follow how adversaries are adapting to this technology, and where it touches your stack today.
Track This Risk
Mallory platform

See the full picture, correlated to your attack surface.

This page covers what’s public. Mallory adds the parts that aren’t: which of your assets are affected, which threat actors are using it right now, and what to deploy next.

Exposure mappingMap indicators from this story to your assets and identify affected systems in minutes.
Threat actor evidenceEvery observed campaign, victim, and pivot linked to actors named in this story.
Associated malwareMalware, exploits, and IOCs connected to the activity described here.
Detection signaturesYARA, Sigma, and Snort rules deployed to your SIEM as soon as they’re published.
Scheduled alertsGet matching new stories delivered to your team as they break, not the next morning.
AI threadsAsk questions about this story and take action on the answers.