TEEs Won’t Secure Agentic AI’s Real-World Interactions
A lot of people think the solution to “private AIs” is to just TEEs. This is already the approach being deployed by Meta, Apple and Google. I think that’s important, but not really a solution. The problem is that for agentic AI, agents need to interact with the real world.
Meta Drops Instagram DM End‑to‑end Encryption
Meta appears to be reversing its strong stance on encryption. The first obvious casualty is that they’re abandoning and disabling end-to-end encryption in Instagram DMs.
EU Advances Message Scanning, Yet Dark Timeline Persists
The EU seems to be going in the right direction when it comes to mass message scanning. Unfortunately, the fact that this vote was necessary proves that we’re still in the dark timeline. https://t.co/Tci3GoLsPg
TikTok Denies “Controversial” Tech, Actually Uses End‑to‑end Encryption
TikTok announces that they’re not going to deploy “controversial privacy tech” that’s actually the same end-to-end encryption most other providers use to protect users’ DMs. https://t.co/INKzu9ku2z
Building Anonymous Credentials for Future Age Verification
New post on anonymous credentials and how to build them. All of this is in service on a longer future post on how these will fit into age verification systems. https://t.co/IqwlP9S6mv
Cloud Providers Can’t Certify TEE Key Secrecy
I heard an interesting anecdote about TEEs from some fintech people. They were trying to convince regulators that TEEs aren’t just “computers under their control”, so they asked cloud providers to certify that they’d never hand over the keys. Providers...
AI Finds FFmpeg Overflow; Patience Beats Complexity
My son showed me one of the overflow vulnerabilities found in FFmpeg by Google/Deepmind’s security AI agents. I was thinking about how hard these things are to find, and at least this one didn’t seem deep — just required enormous...