The Rollup Why The AI Bubble Is Set To Explode (Soon) - Hang Yin
Oct 5, 2025
Hang Yin, co-founder of Phala and a privacy-centered AI researcher, shares crucial insights into the world of confidential computing. He breaks down the latest Intel TEE exploits and discusses why some encryption methods are often overlooked. Yin emphasizes the importance of full-stack confidentiality and decentralized key management for secure AI environments. The conversation furthers into the significance of private AI, comparing its potential impact to that of nuclear fusion, urging everyone to take an interest in privacy solutions.
AI Snips
Chapters
Transcript
Episode notes
Legacy TEE Exploits Don’t End Confidential Computing
- The recent Intel TEE exploits targeted legacy hardware and combined multiple vulnerabilities to succeed.
- Hang Yin emphasizes newer hardware and firmware patches mitigate these specific attacks.
Encryption In Use Is The Hard Problem
- Encryption in transit and at rest are common but encryption in use is the hardest and most neglected.
- Confidential computing aims to protect data during computation so GPUs and CPUs never expose plaintext to operators.
Fala’s Origin: From Blockchain Trust To AI Privacy
- Fala started six years ago to solve trust problems seen in blockchain and applied that mindset to AI.
- They shifted focus to confidential computing and created D‑Stack as an open source solution.