Hackaiz Academy · AI and agents

OWASP LLM Top 10

Main risks of applications using language models.

Automated technical check — neither certification nor legal advice.A service operated by HACKTUALIZ Inc.
  • Prompt injection
  • Sensitive information disclosure
  • Excessive agency
  • Exposed AI keys and unguarded model calls in the repository
  • No active prompt-injection testing of your service

Not actively; it analyses code and configuration.

https://genai.owasp.org/llm-top-10/

Control points

7 automatic · 1 assisted (your answers or documents) · 2 to be proven by a person

  • LLM01 Prompt injectionAutomatic
  • LLM02 Sensitive information disclosureAutomatic
  • LLM03 Supply chainAutomatic
  • LLM04 Data and model poisoning: training data provenanceHuman
  • LLM05 Improper output handlingAutomatic
  • LLM06 Excessive agencyAutomatic
  • LLM07 System prompt leakage: no secrets in instructionsAssisted
  • LLM08 Vector and embedding weaknessesAutomatic
  • LLM09 Misinformation: human review of important answersHuman
  • LLM10 Unbounded consumptionAutomatic

Check your project: domain ownership proof and consent required, each check is billed in credits by actual work. Results are shown as verified points with evidence, never as a certification.