No-LLM Rules
Official Rules
The "No-LLM" rules are defined in point 20 and its subpoints of the Rules and Regulations:
20. Teams may, of their own volition, take part in the “No-LLM” bonus category. By participating in the category, a team undertakes not to use tools and technologies based on large language models (LLM).
20.1. Teams participating in the category are highlighted in CTFd as participating without using LLM tools.
20.2. Teams participating in the category must not, in the scope of the CTF:
20.2.1. use agentic LLM tools (e.g., Codex, Claude Code);
20.2.2. use LLM-based code generation and intelligent completion tools (e.g., Visual Studio Copilot, Cursor);
20.2.3. use LLM-based chat platforms (e.g., ChatGPT, Gemini, Claude);
20.2.4. use LLM tools in any other way.
20.3. A participant in a “No-LLM” team must, after submitting a flag, also submit a short description or the steps of how the challenge was solved.
20.4. If suspicion arrises of a team using LLM-based tools, the organizers retain rights to perform a check and request proof of how the challenge was solved. The organizers retain rights to disqualify a team from the CTF or remove a team from the category, if unauthorized use of LLMs is established.
What does this mean?
If a team participates in the "No-LLM" bonus category, it means they cannot use any LLM tools and technologies to solve challenges.
If the question: "can I use X?" pops up - ask the organizers.
The No-LLM category is meant for teams who want to test and develop their skills using traditional methods. This is understandable, as a challenge and as an opportunity to push yourself to learn something deeper, not just at a surface-level.
Nowadays, avoiding LLMs is almost impossible, so we're asking teams who have undertaken this, to avoid LLMs in places like Google's "AI Overview" and DuckDuckGo's "Search Assist".
What do I need to do for "No-LLM"?
When solving tasks, you will definitely remember what you did to get to the solution, as well as keep some solution scripts, commands, and intermediate results.
When submitting your flag, there will be a solution submission form. You can provide files, which prove how you solved the challenge, as well as a short description of the solve steps, ideas which you thought of during the solving process, and overall thoughts about the challenge.
This description need not be detailed, nor long, nor even gramatically correct. Something like "Looked at the photo's EXIF data" would also work - short and to the point, explains what was done. This really needn't be anything to stress about, insofar the organizers know, what was done, and that it was done by your own hands.
Why should I join this category?
We know that LLMs can solve tasks very fast and in instances even without human supervision. The goal of Mārtiņa-CTF has and will never be participation in order to win, but rather cybersecurity skill and knowledge development and the popularization of cybersecurity, which is why we urge participants to join the "No-LLM" category - to learn and grow, not win by any means necessary.