AI security readiness: AI protection and threats
Podcast|5 November 2026
AI can help your business move faster - but without the right safeguards, it can also expose sensitive data, misuse powerful permissions or make damaging changes at machine speed.
In this episode of Grey Matter Talks Tech, our cyber security expert, Scott Harrison explores the Protect layer of AI security: the guard rails, controls and recovery plans that help prevent incidents and limit their impact.
What you’ll learn
This practical episode explains:
- The four layers of AI protection: identity, data, AI and recovery
- Lessons from real-world AI incidents involving excessive permissions and stolen access tokens
- Six achievable protection controls for SMBs and SMEs
- Why critical AI actions should require human approval
- How immutable, tested backups provide an essential “undo button”
- How ESET, Acronis, Sophos, CrowdStrike and Veeam support different parts of an AI security strategy
The key message is simple: instructions are not controls. Telling an AI agent not to perform a risky action is no substitute for limiting its permissions, approving its connections and ensuring that important decisions remain under human oversight.
Protection does not need to mean blocking innovation. Well-designed controls make the secure option the easiest option - helping employees use approved AI tools confidently while keeping valuable systems and information within safe boundaries.
Listen now - explore more about AI protection and threats
You can tune in via your preferred podcast platform - Acast, Apple, Pocket Casts, or press play on the Spotify link below.
And don't forget to subscribe, so you don't miss out on other great tech insights and best practices.