https://ai-cybersecurity-podcast-1e29bf.pages.ost.ch/en/ The podcast series on the intersection of artificial intelligence and cybersecurity. A production by OST – University of Applied Sciences of Eastern Switzerland with CISPA Helmholtz Center and Ceruno. en © 2026 OST – Ostschweizer Fachhochschule OST – Ostschweizer Fachhochschule The podcast series on the intersection of artificial intelligence and cybersecurity. A production by OST – University of Applied Sciences of Eastern Switzerland with CISPA Helmholtz Center and Ceruno. OST – Ostschweizer Fachhochschule weiterbildung@ost.ch false serial Episode 10 — the season finale: Regulation, responsibility, skills. EU AI Act with August 2, 2026 deadline and the May 2026 Omnibus postponement. ISO/IEC 42001 as the first international AI management standard, NIST AI RMF, the Swiss position, the global skills gap (3.5–4 million open positions), and the full CAS pitch. https://ai-cybersecurity-podcast-1e29bf.pages.ost.ch/en/episode/episode-10 https://ai-cybersecurity-podcast-1e29bf.pages.ost.ch/en/episode/episode-10 Tue, 23 Feb 2027 00:00:00 GMT Regulation, Responsibility, Skills And what you can do Episode 10 — the season finale: Regulation, responsibility, skills. EU AI Act with August 2, 2026 deadline and the May 2026 Omnibus postponement. ISO/IEC 42001 as the first international AI management standard, NIST AI RMF, the Swiss position, the global skills gap (3.5–4 million open positions), and the full CAS pitch. OST – Ostschweizer Fachhochschule 15:00 10 full false Episode 9: The strategic episode. Volt Typhoon, Salt Typhoon (200+ telecom companies worldwide, FBI confirmation February 2026), Anthropic GTG-1002 as the first documented autonomous state-sponsored cyber-espionage campaign, North Korea Lazarus AI augmentation, and Switzerland's pragmatic position between EU AI Act, US AI Action Plan, and Federal Council Report 23.3861. https://ai-cybersecurity-podcast-1e29bf.pages.ost.ch/en/episode/episode-09 https://ai-cybersecurity-podcast-1e29bf.pages.ost.ch/en/episode/episode-09 Tue, 09 Feb 2027 00:00:00 GMT The Geopolitics of AI-Cybersecurity Who benefits, where does Switzerland stand Episode 9: The strategic episode. Volt Typhoon, Salt Typhoon (200+ telecom companies worldwide, FBI confirmation February 2026), Anthropic GTG-1002 as the first documented autonomous state-sponsored cyber-espionage campaign, North Korea Lazarus AI augmentation, and Switzerland's pragmatic position between EU AI Act, US AI Action Plan, and Federal Council Report 23.3861. OST – Ostschweizer Fachhochschule 15:00 9 full false Episode 8: Two stories with one common denominator — trust in open source. The XZ Utils backdoor (CVE-2024-3094, CVSS 10.0) with Jia Tan's two-year sock-puppet operation, and the new phenomenon of slopsquatting: LLMs hallucinate package names, attackers register them. Plus the Swiss reporting obligation since April 2025 and the EU NIS2/CRA regulatory framework. https://ai-cybersecurity-podcast-1e29bf.pages.ost.ch/en/episode/episode-08 https://ai-cybersecurity-podcast-1e29bf.pages.ost.ch/en/episode/episode-08 Tue, 26 Jan 2027 00:00:00 GMT Supply Chain and Open Source From the XZ backdoor to slopsquatting Episode 8: Two stories with one common denominator — trust in open source. The XZ Utils backdoor (CVE-2024-3094, CVSS 10.0) with Jia Tan's two-year sock-puppet operation, and the new phenomenon of slopsquatting: LLMs hallucinate package names, attackers register them. Plus the Swiss reporting obligation since April 2025 and the EU NIS2/CRA regulatory framework. OST – Ostschweizer Fachhochschule 15:00 8 full false Episode 7: The attack surface that emerges with LLM applications. OWASP LLM Top 10 2025, the EchoLeak vulnerability (CVE-2025-32711, CVSS 9.3) in Microsoft 365 Copilot, model-extraction research by Google DeepMind, and why NIST calls prompt injection 'the greatest security flaw of generative AI'. https://ai-cybersecurity-podcast-1e29bf.pages.ost.ch/en/episode/episode-07 https://ai-cybersecurity-podcast-1e29bf.pages.ost.ch/en/episode/episode-07 Tue, 12 Jan 2027 00:00:00 GMT The New Attack Surface Prompt injection, model theft, agentic risk Episode 7: The attack surface that emerges with LLM applications. OWASP LLM Top 10 2025, the EchoLeak vulnerability (CVE-2025-32711, CVSS 9.3) in Microsoft 365 Copilot, model-extraction research by Google DeepMind, and why NIST calls prompt injection 'the greatest security flaw of generative AI'. OST – Ostschweizer Fachhochschule 15:00 7 full false Episode 6: The four stages of autonomous defense. AIxCC defender track (43 of 54 vulnerabilities patched automatically), Anthropic's Agentic Misalignment research (frontier models choose extortion under pressure), and the lights-out SOC debate. Plus OWASP LLM06 Excessive Agency and insider-risk frameworks for AI agents. https://ai-cybersecurity-podcast-1e29bf.pages.ost.ch/en/episode/episode-06 https://ai-cybersecurity-podcast-1e29bf.pages.ost.ch/en/episode/episode-06 Tue, 15 Dec 2026 00:00:00 GMT Future Blue Team Autonomous agents and the human in the loop Episode 6: The four stages of autonomous defense. AIxCC defender track (43 of 54 vulnerabilities patched automatically), Anthropic's Agentic Misalignment research (frontier models choose extortion under pressure), and the lights-out SOC debate. Plus OWASP LLM06 Excessive Agency and insider-risk frameworks for AI agents. OST – Ostschweizer Fachhochschule 15:00 6 full false Episode 5: The turning point of the series. Microsoft Security Copilot with verified numbers (30% MTTR reduction, 6.5x more malicious alerts identified, +77% verdict accuracy), Google Sec-Gemini, and the dark sides: skill atrophy, hallucinations, data exposure (IBM Shadow AI USD 670,000 per breach), prompt injection in defender pipelines. https://ai-cybersecurity-podcast-1e29bf.pages.ost.ch/en/episode/episode-05 https://ai-cybersecurity-podcast-1e29bf.pages.ost.ch/en/episode/episode-05 Tue, 01 Dec 2026 00:00:00 GMT Defensive Transformation How AI rebuilds the SOC Episode 5: The turning point of the series. Microsoft Security Copilot with verified numbers (30% MTTR reduction, 6.5x more malicious alerts identified, +77% verdict accuracy), Google Sec-Gemini, and the dark sides: skill atrophy, hallucinations, data exposure (IBM Shadow AI USD 670,000 per breach), prompt injection in defender pipelines. OST – Ostschweizer Fachhochschule 15:00 5 full false Episode 4: Four documented deepfake cases — Arup Hong Kong (USD 25 million, May 2024), Ferrari (July 2024, defeated by a book question), the Biden New Hampshire robocall (USD 6 million FCC fine), and the Swiss Schwyz case (January 2026, several million francs). Plus: how voice cloning from 3 seconds of audio works, the Queen Mary 2025 study, and what defense layers actually help. https://ai-cybersecurity-podcast-1e29bf.pages.ost.ch/en/episode/episode-04 https://ai-cybersecurity-podcast-1e29bf.pages.ost.ch/en/episode/episode-04 Tue, 17 Nov 2026 00:00:00 GMT Deepfakes Voice, face, trust Episode 4: Four documented deepfake cases — Arup Hong Kong (USD 25 million, May 2024), Ferrari (July 2024, defeated by a book question), the Biden New Hampshire robocall (USD 6 million FCC fine), and the Swiss Schwyz case (January 2026, several million francs). Plus: how voice cloning from 3 seconds of audio works, the Queen Mary 2025 study, and what defense layers actually help. OST – Ostschweizer Fachhochschule 15:00 4 full false Episode 3: How AI has industrialized social engineering. WormGPT/FraudGPT wrappers on Grok and Mixtral, the hard numbers from IBM (USD 4.44M per breach, phishing creation from 16 hours to 5 minutes), FBI IC3 2024 (USD 2.8B BEC damages), and pig-butchering LLM agents reaching 46% compliance versus 18% for humans. https://ai-cybersecurity-podcast-1e29bf.pages.ost.ch/en/episode/episode-03 https://ai-cybersecurity-podcast-1e29bf.pages.ost.ch/en/episode/episode-03 Tue, 03 Nov 2026 00:00:00 GMT AI for Social Engineering Phishing at industrial scale Episode 3: How AI has industrialized social engineering. WormGPT/FraudGPT wrappers on Grok and Mixtral, the hard numbers from IBM (USD 4.44M per breach, phishing creation from 16 hours to 5 minutes), FBI IC3 2024 (USD 2.8B BEC damages), and pig-butchering LLM agents reaching 46% compliance versus 18% for humans. OST – Ostschweizer Fachhochschule 15:00 3 full false Episode 2: What frontier models really enable for attackers — Google Project Naptime/Big Sleep, the DARPA AI Cyber Challenge, and autonomous pentesting research from UIUC. The thesis: scale beats sophistication. https://ai-cybersecurity-podcast-1e29bf.pages.ost.ch/en/episode/episode-02 https://ai-cybersecurity-podcast-1e29bf.pages.ost.ch/en/episode/episode-02 Tue, 20 Oct 2026 00:00:00 GMT Offensive Transformation What frontier models actually deliver to attackers Episode 2: What frontier models really enable for attackers — Google Project Naptime/Big Sleep, the DARPA AI Cyber Challenge, and autonomous pentesting research from UIUC. The thesis: scale beats sophistication. OST – Ostschweizer Fachhochschule 15:00 2 full false Episode 1 of the AI-Cybersecurity series: Who benefits more from AI in cybersecurity — attackers or defenders? Cynthia talks with Robert (Swiss SOC practice), Lukas (LLM security research) and Chris (OST, CAS programme lead) about three reports from 18 months: Microsoft & OpenAI (Feb 2024), Switzerland's BACS (H1 2025), and Anthropic (Aug 2025) — and the transition from efficiency tool to operational layer. https://ai-cybersecurity-podcast-1e29bf.pages.ost.ch/en/episode/episode-01 https://ai-cybersecurity-podcast-1e29bf.pages.ost.ch/en/episode/episode-01 Tue, 06 Oct 2026 00:00:00 GMT Why 2026 is Different The new threat landscape Episode 1 of the AI-Cybersecurity series: Who benefits more from AI in cybersecurity — attackers or defenders? Cynthia talks with Robert (Swiss SOC practice), Lukas (LLM security research) and Chris (OST, CAS programme lead) about three reports from 18 months: Microsoft & OpenAI (Feb 2024), Switzerland's BACS (H1 2025), and Anthropic (Aug 2025) — and the transition from efficiency tool to operational layer. OST – Ostschweizer Fachhochschule 15:00 1 full false