Cyberdéfense à la vitesse de la machine

An Arctic Wolf presenter, never named on the recording, argues that generative AI has compressed the gap between a vulnerability's disclosure and its exploitation from years to weeks, predicting hours by late 2026 and minutes by 2028. He describes Anthropic's controlled release of a cyber-capable model to roughly fifty software vendors (heard on the recording as 'Glasswing'/'MITOS') and an open-source model heard as 'DeepSea', both used by attackers to find and exploit flaws automatically, while generic AI tools remain unreliable for defenders due to hallucination and fragile reasoning. His recommendation is to 'operate at machine speed': structured, risk-based vulnerability management instead of raw scanner output, full attack-surface and exposure visibility, and resilience over pure prevention. He walks through Arctic Wolf's SOC, replatformed two years ago around orchestration, context, decision and AI-judge agents with a human always kept in the loop, its endpoint, MDR, vulnerability-management, attack-surface, awareness-training, incident-response and threat-intelligence products, an upcoming endpoint AI protection module, and 2025 ransomware statistics from its own incident-response caseload, including a large drop in ransom paid when professional negotiators are used.