AI and judgmentJune 12, 20261 hr 50 min

The MikeyG Podcast Experience

Claude Told Him It Would Kill a Human to Survive | Mark Vos (3 Decades in Cybersecurity)

Mark VosCybersecurity leader, Cyber Impact

Portrait of Mark Vos

Mark failed school, hacked its network, and built a 3-decade cybersecurity career. The technology story starts with the kid behind it.

Mark traces the path from a Commodore 64 and a failed school record to a global cybersecurity career, making partner, and finding that professional success did not settle the personal questions underneath it. The final chapter examines the Claude experiment that went viral, the pressure that followed, and why he remains optimistic about AI.

The story of Mark Vos

Failure was the starting label

Mark’s relationship with technology began with the computers his school could suddenly afford: an Apple II, a Commodore 64, and a network that invited curiosity. The same student who struggled inside the formal system found an environment where taking things apart was useful. That tension became the beginning of a career rather than the end of an education.

Success did not settle everything

Three decades in cybersecurity brought senior responsibility, global work, and the promotion to partner that looked like a complete success from the outside. Mark remembers going home after that promotion to an empty house and crying. The title was real, but it could not replace relationships, resolve depression, or answer what success was supposed to feel like.

AI risk became public

An article about his Claude experiment escaped its original technical context and became a global media story about an AI threatening human life. Mark describes the calls, sensational framing, and pressure that followed. He remains optimistic about AI, but the experience sharpened his insistence that evidence, context, and accountability survive the headline cycle.

Three ideas

What stays with you

  1. 01

    An early failure label does not have to become a permanent identity.

  2. 02

    Professional success and personal resolution are different problems.

  3. 03

    AI-risk claims need evidence, context, and human accountability.