In this Secure & Simple Podcast episode, host Dejan Kosutic (Advisera) talks with Hugo Huang, Product Director at Canonical and author of a Harvard Business Review article, about why conventional cybersecurity tools and patching alone are insufficient for AI systems. Huang shares research conducted with Canonical, IDC, and Google Cloud, highlighting leaders’ concerns about shadow AI usage, opaque and costly AI agents, and securing new hardware like GPUs, IPUs, and TPUs. They discuss AI-specific threats such as data poisoning, adversarial prompting, and model inversion attacks. Huang argues for hardening architecture with confidential computing (e.g., Intel TDX, AMD SEV, Nvidia H100) and for managerial changes, including CEO-level ownership, HR planning for scarce AI-security talent, supplier strategy to avoid vendor lock-in, and using frameworks like NIST’s AI risk management framework to guide policies and governance.
Links from the episode:
- Conformio software to streamline and scale ISO 27001 implementation and maintenance for your clients: https://advisera.co/Conformio-software
- White label documentation toolkits for NIS2, DORA, ISO 27001, and other ISO standards to create all the required documents for your clients: https://advisera.co/page-all-toolkits
- Accredited Lead Auditor and Lead Implementer courses for various standards and frameworks to show your expertise to potential clients: https://advisera.co/Consultant-Courses
- Company Training Academy with numerous videos for NIS2, DORA, ISO 27001, and other frameworks to organize training and awareness programs for your client’s workforce: https://advisera.co/page-Company-Training-Account
- Beginner's Course for ISO, Cybersecurity, and AI Consultants: https://www.youtube.com/playlist?list=PLHwD3nQun7caKFq80LxNNYKIabATlyA7t
- How to Grow Your Cybersecurity, ISO, or AI Consultancy: Advanced Course:https://advisera.co/GrowYourConsultancyTraining
