AI Security Watch

Automated database of LLM/ML vulnerabilities and incidents

Automatic collection active
-
Vulnerabilities
-
Critical
-
Last 24h
4+
Sources
NVD/NIST (CVE)
arXiv (Papers)
GitHub Security
10+ security blogs

Last update: loading...

|

Detected vulnerabilities

Loading...

Loading vulnerabilities...

OWASP LLM Top 10 (2025)

The 10 most critical security risks for LLM applications, per the OWASP Foundation.

1

Prompt Injection

User input or external content alters model behavior, directly or through indirect injection.

2

Sensitive Information Disclosure

Personal, financial or business data leaked through model responses or memorization.

3

Supply Chain Vulnerabilities

Risks from foundation models, hosted APIs, fine-tunes, RAG sources and third-party tools.

4

Data and Model Poisoning

Training or fine-tuning data manipulated to implant bias or backdoors.

5

Improper Output Handling

Model output passed downstream without validation or sanitization.

6

Excessive Agency

Too much functionality, permission or autonomy granted to an agent.

7

System Prompt Leakage

System instructions disclosed, exposing rules, secrets or architecture.

8

Vector and Embedding Weaknesses

Flaws in how embeddings are generated, stored or retrieved in RAG systems.

9

Misinformation

Credible-looking false output, including hallucinated facts and code.

10

Unbounded Consumption

Unrestricted inference driving denial of service, cost blow-up or model extraction.