Marking Your Own Homework

Why Human-Facing AI Needs an Outside Examiner

Most people assume AI safety checks are independent, but the product's makers usually run them. More than 15 years of ventures where technology meets people makes the outside check matter most. Who would you trust to grade an AI product used by someone in a crisis? Shannon learned that an outside examiner protects the builders as much as the people who use it. Ask any AI product that claims to be safe what its outside check tested and who ran it. Grading your own work is quicker; an independent check finds what the builders missed. Get an outside examiner before launch, and users stop being the test.

About Shannon Harvey

Shannon Harvey, podcast guest

Founder & AI Safety Researcher

15+ years Founding and leading studios and research initiativesApple TV+ Virtual production leadership on Silo and HijackProduction Park Established its research labReflection Labs Founder, independent governance for human-facing AI

Shannon Harvey is a founder, researcher and innovation leader focused on AI safety, human behaviour and responsible technology. He is currently developing research and practical testing frameworks to help companies identify and reduce the emotional, psychological and social risks created by human-facing AI.

With 15+ years of experience spanning technology, immersive media, live production, research and academia, Shannon brings a unique perspective shaped by both systems engineering and a background in performance. He has led complex technical and creative projects, including virtual production for Apple TV+ productions such as Silo and Hijack, research at Production Park, and platform adoption work with disguise. He was also Course Leader for the BA (Hons) Visual Production degree at Backstage Academy, now the Academy of Live Technology.

Shannon originally trained and performed as an opera singer before moving into theatre, and large-scale live production. His experience of performance, audience interaction and safety engineering has shaped the way he thinks about AI: particularly the importance of understanding context, vulnerability, trust and what can happen when systems behave unexpectedly.

His current research explores the often-overlooked side of AI safety - what happens when people trust, depend on or emotionally engage with AI systems, particularly in high-trust and vulnerable situations.

View profile →

Podcast Hosts: Get Shannon’s Media Profile

Click to Download