0x34C - Teknik - Série l'IA dans les CtF - FCSC

0x34C - Teknik - Série l'IA dans les CtF - FCSC

0x34C - Technique - The AI in CTFs series - FCSC

by PolySécure Podcast · View the channel on PSI

Humanities, Social Sciences & Thought AI — News & Applications UYComputer scienceUYQArtificial intelligence (AI)
🎙 PolySécure Podcast 👥 552 📅 October 9, 2026 ⏱ 39 min 👁 4 📄 expert opinion 🧭 2026-10-09
Available in: English (current) Français

Keywords

CTFAIANSSIcheatingcompetition

Summary

The interview discusses the profound impact of large language models (LLMs) on Capture The Flag (CTF) competitions, particularly the FCSC organized by the French cybersecurity agency ANSSI. Jérémy Jean, a cryptologist at ANSSI, explains that LLMs have become extremely effective at solving CTF challenges, drastically altering the competitive landscape. In 2025, models were hallucinating flags, but by 2026 they could solve challenges in minutes, rendering months of design work trivialized. This forces organizers to ban LLM use, leading to a difficult policing effort during the 10-day event. They used tickets and suspected participants to explain their methods. About half denied cheating, but hidden accounts helped maintain a credible scoreboard. The selection process for the French team for the ECSC involved interviews with top players. The ECSC itself is considering similar bans, especially in a face-to-face format. The video also highlights Accropole, a non-competitive platform with over 600 past FCSC challenges, as a stress-free learning alternative. Organizers are demoralized, and the future of CTFs is uncertain, with possibilities ranging from new rules to discontinuation. The conversation touches on parallels with doping in sports and the need to rethink the purpose of competitions in the age of AI.

196 words

Critical Evaluation

Value of the Information & Strength of the Argument

The interview provides valuable first-hand insight from an organizer’s perspective, detailing the concrete challenges posed by LLMs in competitive cybersecurity. Jean’s argumentation is solid, built on specific examples such as a reverse engineering challenge solved in minutes after weeks of design. He effectively illustrates the unintended consequences of LLM usage, including the loss of learning value and the breakdown of fairness. The reasoning is coherent and avoids overgeneralization, acknowledging uncertainty about future solutions. The discussion of Accropole as a non-competitive alternative adds practical value. However, the reliance on anecdotal evidence and the absence of broader data (e.g., statistics on cheating rates) slightly weakens the overall argumentative foundation.

Scientific Rigor, Source Quality, Title Accuracy

The video maintains a good level of scientific rigor by relying on the speaker’s direct involvement in ANSSI and FCSC. The only external reference is the Accropole platform, which is mentioned by name and URL, and it serves as a credible resource. The title is appropriate as it directly maps to the content, focusing on AI in CTFs with an emphasis on the FCSC. No external sources are cited beyond the platform, and the discussion is based on practical experience rather than published literature. The quality of the information is high given the expert status of the guest, though the lack of verifiable citations limits its technical depth. The adequacy between title and content is strong, with no significant deviation.

242 words

Title / Content Match

The title accurately reflects the content, which focuses on the role of AI (LLMs) in CTF competitions, specifically the FCSC.

Quality & Reliability

7/10

The video features an expert from ANSSI sharing direct experience on LLM impact in CTFs. Information is credible but based on personal experience and anecdotal evidence, with no external references cited. The discussion is coherent, detailed, and provides practical insights.

Key Moments

Markers derived by PSI from the transcript: the creator did not define chapters.

Cited Sources

  • Accropole — Mentioned as a non-competitive platform with over 600 past FCSC challenges, available for self-training.

Contribution & Novelties

The video provides a detailed, first-hand account of how a major national CTF organizer (ANSSI) copes with LLM-driven cheating, offering insights into practical countermeasures and the psychological toll on organizers. It also introduces Accropole as a non-competitive complementary resource, a novel approach to preserve learning value in an AI-dominated era.

Pour aller plus loin :

  • Capture the flag (cybersecurity) — Provides background on CTF competitions and their formats.
  • Large language model — Explains the technology behind LLMs and their capabilities.
  • ANSSI — Official agency page, offering context on the French cybersecurity agency’s role and activities.

95 words

Radar Profile

The radar profile shows strong scores in quality and quantity of information (7-8), indicating a content-rich interview. Technical level is moderate (6), suitable for a general audience. Fiabilité is high (7) due to the expert status, though the lack of external references prevents a higher score. Overall, the content is well-balanced, with a slight lean towards qualitative depth rather than technical complexity.

Reliability 7/10