threatdeck
Live 13:45 UTC ·

Learn / AI security fundamentals

Lesson 3 of 510 minReviewed

The OWASP LLM and Agentic Top 10 lists, in practice

Know what each entry of the two OWASP lists means for an engineer, how the lists relate, and how to use the ids in threat models, findings and detections without treating them as a checklist.

The OWASP GenAI Security Project maintains two community lists. The Top 10 for LLM Applications, 2025 edition, covers applications built on large language models [1]. The Top 10 for Agentic Applications for 2026, published in December 2025, covers systems where models plan and act through tools, memory and other agents [2]. Neither is a standard you comply with; both are a shared vocabulary for the failures practitioners keep seeing.

OWASP Top 10 for LLM Applications 2025 (ThreatDeck's one-line reading)
EntryWhat goes wrong
LLM01 Prompt InjectionDirect or indirect input changes the model's behaviour against the developer's intent
LLM02 Sensitive Information DisclosureSecrets, personal or proprietary data come back in output
LLM03 Supply ChainCompromised models, datasets, packages or plugins
LLM04 Data and Model PoisoningManipulated training, fine-tuning or embedding data
LLM05 Improper Output HandlingOutput reaches a browser, shell or query without validation
LLM06 Excessive AgencyMore functionality, permission or autonomy than the task needs
LLM07 System Prompt LeakageSecrets or logic placed in prompts get extracted
LLM08 Vector and Embedding WeaknessesRAG leaks across users or ingests poisoned content
LLM09 MisinformationConfident false output, including hallucinated packages
LLM10 Unbounded ConsumptionRunaway usage, cost or model extraction

The agentic list exists because autonomy changes the failure modes. When a model chooses tools, keeps memory and delegates to other agents, the question shifts from what it says to what it does and with whose authority [2]. OWASP's earlier Agentic AI Threats and Mitigations guide (February 2025) laid the threat-model groundwork for it [3].

OWASP Top 10 for Agentic Applications 2026
EntryWhat goes wrong
ASI01 Agent Goal HijackInjected content redirects the agent's objective or plan
ASI02 Tool Misuse and ExploitationLegitimate tools used destructively or to exfiltrate
ASI03 Identity and Privilege AbuseAgents inherit or escalate credentials beyond the task
ASI04 Agentic Supply Chain VulnerabilitiesMalicious MCP servers, skills, plugins or prompts
ASI05 Unexpected Code ExecutionGenerated code runs outside the intended sandbox
ASI06 Memory and Context PoisoningPersistent memory seeded to influence later sessions
ASI07 Insecure Inter-Agent CommunicationAgent messages spoofed or tampered with
ASI08 Cascading FailuresOne faulty agent propagates through connected workflows
ASI09 Human-Agent Trust ExploitationPeople over-trust agent output or approve blindly
ASI10 Rogue AgentsAgents act outside scope or against their principals

How the lists relate

Several pairs describe the same root cause at different layers. Prompt injection (LLM01) is the usual mechanism behind goal hijack (ASI01). Excessive agency (LLM06) is what turns a hijack into tool misuse (ASI02) or privilege abuse (ASI03). Supply chain appears in both, but ASI04 adds what agents load at run time: MCP servers, skills and tool descriptions that can change after you approved them. Use the agentic id when the harm comes from an action, and the LLM id when it stays in the model's input or output.

Using the ids well

  • Threat models. Tag each row with one or two ids so reviewers can compare systems and spot gaps, such as a design with no LLM05 analysis even though output is rendered as HTML.
  • Findings. Put the id in the title of red-team findings next to the concrete impact. The id says which class; the impact says how bad.
  • Detections and tests. ThreatDeck's detections and the coverage tracker carry the same ids, so you can see which risks you test and which you only document.
  • Not a checklist. A system can be clean on all twenty entries and still fail on its own business logic. The lists say where to look first, not where to stop.

For governance teams, NIST's Generative AI Profile maps the same ground to risk-management actions [4]. For attack sequences and detection mapping, the next lesson covers MITRE ATLAS, which the technique pages on ThreatDeck cross-reference with these ids.

Try it on ThreatDeck

Self-check

4 questions

Pick an answer to see whether it is right and why. Only your score is saved, in this browser.

  1. An injected web page makes a coding agent run a destructive cloud command. Which pair of ids best describes the mechanism and the harm?
  2. Why did OWASP publish a separate Top 10 for Agentic Applications?
  3. What is the right way to use the lists in a security review?
  4. A RAG assistant shows one customer another customer's documents. Which entry fits best?

References

  1. OWASP Top 10 for LLM Applications 2025. OWASP GenAI Security Project, 2024-11.
  2. OWASP Top 10 for Agentic Applications for 2026. OWASP GenAI Security Project, 2025-12.
  3. Agentic AI – Threats and Mitigations. OWASP GenAI Security Project, 2025-02.
  4. NIST AI 600-1: Artificial Intelligence Risk Management Framework: Generative Artificial Intelligence Profile. NIST, 2024-07.
Completing the self-check also marks the lesson complete. Saved in this browser only.

Following

Follow the vendors, products, AI tools, techniques and actors you care about. Matching stories, CVEs and exploited bugs are marked, a Following tab appears in the feed, and returning visits show what changed. Saved only in this browser.

Presets

Get updates

A personal RSS feed of new stories on everything you follow. Paste it into any feed reader, Slack or Teams RSS app.