Cyber Heard

No Defences: The Honest State of AI Security

Chris Jefferson, Co-Founder & CTO at Advai

16 September 2026 · 24 min

Chris Jefferson explains why AI systems fail by design rather than by accident, where language models and vision systems break under attack, and why some of these weaknesses have no real defence. He also covers the practical risks of data leakage, AI generated code, and handing tasks to autonomous agents.

AI SecurityAdversarial TestingAI RegulationData LeakageAgentic AI

Listen and subscribe

Spotify Apple Podcasts YouTube RSS

More episodes

14 September 2026

The Bot Is Another Employee

Nick Proud

9 September 2026

People, Patients and Ransomware: The Real NHS Threat

Daniel Kay

7 September 2026

The Signaling Layer: Telecom's Blind Spot Attackers Love

Dmitry Kurbatov

The 24-hour rule
Why slow Cyber Hiring is now a board risk

A 16-page invitise briefing for cyber, risk and people leaders. Why the hiring window has narrowed, what 24-hour mobilisation actually requires, and the four conditions that have to be true for next-morning on-site to mean something.

55%

of senior cyber roles take 6 months or longer to fill in the UK.

46 days

longer than other IT roles, senior cyber vacancies stay open.

68%

of large UK businesses now own cyber at board level.

No Defences: The Honest State of AI Security

More episodes

The 24-hour rule Why slow Cyber Hiring is now a board risk

The 24-hour rule
Why slow Cyber Hiring is now a board risk