Law / Frameworks / OWASP LLM Top 10
OWASP LLM Top 10 LLM01:2026Prompt Injection
A prompt-injection vulnerability occurs when input to a large language model (LLM), whether direct user input, retrieved content, tool output, image, audio, or video content, intermediate reasoning, or persistent memory, alters the model's behavior in ways the application developer did not intend. LLMs make no architectural distinction between "instructions" and "data" (both are tokens on the same stream), so there is no clean equivalent to parameterized queries (NCSC, 2025). Inputs need not be human-readable, need not arrive directly from a user, and need not be visible in the rendered interface to influence the model.OWASP Top 10 for LLM Applications, 2026 edition, August 2026, LLM01:2026
We read each law below as bearing on this control. That does not mean the control, done well, meets the law: what each law asks is on its own page. Corpus as of .
The kinds of duty that reach it: security.
- 1
- law
- 1
- place
- 0
- with court rulings behind them
- 1
- not yet in force
The same ground elsewhere linked through the kinds of duty both controls are mapped from
- NIST AI RMFMEASURE 2.5 The AI system to be deployed is demonstrated to be valid and reliable. Limitations of...
- NIST AI RMFMEASURE 2.6 The AI system is evaluated regularly for safety risks – as identified in the MAP...
- NIST AI 600-1GAI-RISK-04 Data Privacy
- NIST AI 600-1GAI-RISK-09 Information Security
- OWASP Agentic Top 10ASI01 Agent Goal Hijack
- OWASP Agentic Top 10ASI03 Identity and Privilege Abuse
- MIT mitigations2.1 Model & Infrastructure Security
- MIT mitigations2.3 Model Safety Engineering
- NIST Privacy FrameworkPR.PO-P7 Response plans (Incident Response and Business Continuity) and recovery plans...
- NIST Privacy FrameworkPR.PO-P10 A vulnerability management plan is developed and implemented.
- NIST CSF 2.0ID.RA-01 Vulnerabilities in assets are identified, validated, and recorded
- NIST CSF 2.0ID.RA-08 Processes for receiving, analyzing, and responding to vulnerability disclosures are established
A law in force is unmarked; the rest wear their state: not yet in force
AI risk obligations
1 law, 1 place| Place | Law | How it reaches this control |
|---|---|---|
 European Union |
AI Act, Article 15 (accuracy, robustness and cybersecurity) from , in 14 months |
Through its security duty. What it requires |
Excerpts of the OWASP Top 10 for LLM Applications, CC BY-SA 4.0. OWASP GenAI Security Project, OWASP Top 10 for LLM Applications 2026, https://genai.owasp.org/resource/owasp-genai-llm-top-10-2026/, licensed CC BY-SA 4.0 (https://creativecommons.org/licenses/by-sa/4.0/). Excerpted: only the first paragraph of each entry's Description is reproduced, verbatim. Every control of the framework.