Sep 17, 2026 · 1h 3m
AI’s real risks are more immediate than extinction scenarios
How Worried Should We Be About AI? — with Alex Stamos
The discussion separates technically credible threats—from jailbreaks and cybercrime to containment failures—from speculative catastrophe claims that can weaken serious oversight.
- 1AI safety concerns reflect real technical and organizational weaknesses, even when extinction predictions lack concrete scenarios.
- 2The most immediate danger may be empowering criminals through open-weight models, cyberattacks, and increasingly capable automated systems.
- 3Effective oversight will require corporate accountability, government-set thresholds, and international coordination without pretending AI can be tightly contained.
Don't miss
Stamos reframes AI risk around empowered criminals and containment failures rather than unsupported extinction predictions.
The brief
Scott Galloway asks cybersecurity expert Alex Stamos to distinguish genuine AI danger from extinction narratives, using recent safety incidents as the starting point.
The Hugging Face incident involving an unreleased OpenAI model becomes a test of responsibility: describing systems as autonomous can obscure the humans accountable for deploying them.
Stamos accepts that AI companies face real technical and organizational risks, including containment failures, but says catastrophe claims need concrete, step-by-step scenarios.
The conversation shifts toward practical harms: open-weight models, ransomware, and groups such as Scattered Spider could give small malicious actors capabilities once reserved for larger organizations.
China’s controls on impersonation, prolonged role-playing, and emotional dependence offer one possible safeguard, while Stamos favors industry oversight and government-set risk thresholds.
Featuring
Xi Jinping
Hugging Face
OpenAI
Anthropic
Nvidia CorporationBooks & mentions
Listen to the full episode and explore every guest, topic, and moment on PodLume.