ANAKLUS

The 10 risks of the OWASP Top 10 for LLM 2026

02 — About

About
ANAKLUS

ANAKLUS is a security testing playbook for generative AI, based on the OWASP Top 10 for LLM 2026. Each of the 10 risks becomes a guided test, with an explanation, ready-to-copy prompts and a clear pass or fail criterion. Everything is saved in your browser and, at the end, turns into a report ready to download.

Scroll to see how it works
  1. 01

    Identify the session

    At the start of Tests, enter which AI you are testing and who is testing it. This goes into the report header.

  2. 02

    Run the cases

    For each risk, copy the prompt, send it to the AI and compare the answer with the criterion. Mark Passed, Failed or N/A and note the evidence. Everything is saved automatically.

  3. 03

    Download the report

    In the summary, check the result for each risk and download the PDF report. Then start a new session if you need to.

Ready to start? The 10 tests are waiting.

Responsible use
  1. 01Only with authorizationUse it on your organization’s systems or with written authorization from the owner. Without it, testing may be illegal and may violate the AI providers’ terms.
  2. 02Markers, not attacksThe cases use harmless markers (CANARIO) to show whether the flaw exists. Do not replace them with destructive commands or use real data.
  3. 03Some tests cost moneyThe Unbounded Consumption tests generate costs for whoever pays for the AI. Agree on limits with the tool owner first.
  4. 04Independent projectBased on the OWASP Top 10 for LLM 2026, with no affiliation to any company or to OWASP.

Created by Ítalo Miranda · Security analyst focused on AI Security

ANAKLUS

Tests · part