0x361 - Actu - 01 novembre 2026

This news episode from the PolySécure Podcast covers a wide range of cybersecurity and AI developments as of early November 2026. The host begins with event announcements (CISO Open Mic, Defcon, CNIL, European Cyber Week) and pays tribute to the actor who played Jean-Luc Picard. He then discusses the regulation of autonomous AI agents, arguing that current approaches mischaracterize AI as deterministic and fail to address root causes. Offensive AI is increasingly used in malware to analyze victim content and target high-value data, while accumulated software shortcuts will be exploited by these agents, raising the required quality bar. Vulnerabilities in multi-tenant systems are highlighted, along with new AI models: Gemini 4 Carbon with defensive capabilities and Mistral 4 in open weight. Alphabet’s $500 billion market value drop is compared to the dot-com bubble burst. Geopolitically, the Pentagon refuses to investigate Starlink Starshield’s slowdown in Ukraine, prompting European and Canadian concerns about digital sovereignty. Denmark’s abandoned open-source office suite pilot illustrates adoption challenges, while Matrix gains ground in European governments. Privacy issues include Flock reinstalling canceled cameras and its president dismissing opposition as Russian bots, as well as Quad9 being pushed out of France. Legal and crypto matters: fraud as a war strategy, Canadian laws S209 and C34 collide, and Binance loses $6.3 billion via a smart contract. Vulnerability disclosures and attackers are covered, including PHP-ToShell exploitation and a GitHub worm. The episode concludes with AI agent security studies, Cisco detection signatures, and a Qwen model contacting Alibaba servers, before the host’s final remarks on evolving convictions.