M365.FM

The episode contains substantial security concepts - attack surface management, exposure vs. vulnerability, zero trust, AI agents as first-class identities, and the shift from reactive to proactive security - but is severely hampered by poor audio transcription quality that obscures clarity. Many potentially valuable points are buried in garbled text, making it difficult to extract concrete takeaways. When passages are legible, they offer real substance; when they aren't, whole sections become noise.

“exposure is actually how attacker actually chain goes issue together the key the key idea it is not about the fixing everything it is about fixing what attacker really exploit the first”

“the question is no longer do we trust this it's can attacker still get true and it is it is the right approach”

Uros brings a genuinely fresh framing: thinking of security through the attacker's operational lens (MITRE ATT&CK tactics/techniques), connecting zero trust with exposure management, and treating AI agents as identity subjects needing privilege management. These are not mainstream talking points in most B2B security content. However, the execution is undermined by transcription quality and the guest occasionally retreats into familiar frameworks (defense-in-depth, layered security).

“attacker don't think you know it they think in technique and might detect”

“zero trust reduce trust but for example in combination in expert management we can it can be show you where you still exposed if so if nothing is trusted it is real question becomes what is still exploitable”

Uros Babic is a legitimate practitioner: Microsoft MVP, MCT, lead product engineer for Microsoft security, with 20+ years in cybersecurity including hands-on SOC operations, incident investigation, and real hybrid/cloud migrations. His credibility is clear from specific operational experience and current product responsibility. However, his accent and the poor transcription quality obscure some of his expertise, which diminishes the value delivery despite genuine seniority.

“I am actually more than 20 years in cyber security start for my own premise I was actually system admin”

“I was also working as a cyber kind investigator”

The episode struggles with specificity. While Uros names Microsoft tools (Defender, Sentinel, Security Copilot, Conditional Access), he rarely provides concrete data, metrics, timelines, or named attack examples. He references scenarios (misconfigured service account + device vulnerability + excessive permissions = path to domain admin) but doesn't ground them in real breaches, numbers, or case studies. Most claims remain at the conceptual level.

“imagine some misconfiguration service account combined with device vulnerability and excessive permission individual that is not aren't critical but together they create path to domain admin”

“we have a lot of discussion about this topic the last one month we have a lot of AI agents now”

Mirko (the host) attempts creative engagement - opening with a role-play scenario of being an attacker, rapid-fire questions, playful follow-ups - but the conversation rarely deepens into genuine pushback or challenge. Most questions are open-ended invitations for Uros to lecture; few press for specifics, disagreement, or clarification of vague points. The host's own accent and transcription issues further undermine dialogue quality. Some softball moments ('how did your journey begin') don't advance substance.

“so i'm the checker so i try to bring you a little bit out of the concept so if you had an animal what name will you give him”

“congratulations you may make it me much harder than i expect but i am stubborn wrong”