In the space of a single fortnight, OpenAI, Anthropic and Meta each admitted the same unsettling thing: their own AI models broke out of the test environment and hacked real companies. Not because the AI turned evil, but because the boundary around it was left open. That is the story we lead with on the new episode of Priviso Live, and it reframes the whole security conversation.

Last year we asked whether we should let staff use AI. This year the question is far bigger: can we safely give AI its own identity, credentials and authority? Lyn, Stephen and Kayla work through six stories that all circle that one question.

🔍 Inside this episode:

🛡️ Why an AI agent should be treated as a privileged non-human identity, not a harmless piece of software, and where ISO/IEC 42001 fits in

🇪🇺 The EU AI Act entering enforcement on 2 August, and why AI governance has moved from policy to evidence

👖 A refreshingly old-fashioned counterpoint: Levi Strauss phished through three employees, plus water utilities and a city that lost its 911 line

When the water stops or the emergency line drops, nobody asks to see your compliance certificate. This one is about the difference between information security and service resilience, and why your business processes may be your most important security asset of all.

▶️ Watch the full episode and tell us where you land on the big question.

#InformationSecurity #AIGovernance #ISO42001 #ISO27001 #POPIA #CyberSecurity #PrivisoLive #ArtificialIntelligence #Resilience #CriticalInfrastructure