Joseph Lorenzo Hall, PhD (@JoeBeOne) header
Joseph Lorenzo Hall, PhD (@JoeBeOne) profile picture

Astrophysicist turned advocate: guiding the globe to a safer, stronger Internet. Distinguished Technologist @internetsociety, @joebeone@techpolicy.social

Joined February 2009
Photo by Joseph Lorenzo Hall, PhD (@JoeBeOne)Photo by Joseph Lorenzo Hall, PhD (@JoeBeOne)Photo by Joseph Lorenzo Hall, PhD (@JoeBeOne)Photo by Joseph Lorenzo Hall, PhD (@JoeBeOne)Photo by Joseph Lorenzo Hall, PhD (@JoeBeOne)Photo by Joseph Lorenzo Hall, PhD (@JoeBeOne)Photo by Joseph Lorenzo Hall, PhD (@JoeBeOne)Photo by Joseph Lorenzo Hall, PhD (@JoeBeOne)Photo by Joseph Lorenzo Hall, PhD (@JoeBeOne)Photo by Joseph Lorenzo Hall, PhD (@JoeBeOne)
Well, hello. Bulgarian wine. Had a 2024 Gomotartzi Dimyat & Vermentino blend (from the Danubian Plains) at Equinox on 19th last night; was an absolute delight.
1
147
Dear @digitalocean: the Droplet recovery console has a bug that gets stuck in ALL CAPS, making it impossible to log in as `root` or enter a mixed-case password. Your docs acknowledge this and suggest mashing Shift/Ctrl until it stops. It's a *recovery console*. Seriously?
1
1
1
3
715
great book for horror fans, especially Friday the 13th devotees: Stacie Ponder’s "Death Count: All of the Deaths in the Friday the 13th Film Series, Illustrated." A lovingly illo'd, deranged, wickedly funny tour of the series' body count. amzn.to/48CQQ4W
1
123
Joseph Lorenzo Hall, PhD retweeted
This is really clever. It's not just a distraction, it also actively elicits useful information about the malicious agent. You could imagine how these fake inference APIs could also inject tool calls that the agents may execute...
Amazon set up decoy servers that pretend to be LLM APIs, and attackers' AI agents ended up sending their entire campaign memory straight to them - system prompts, tool histories, the shared knowledge graph and the findings they had logged. Amazon's MadPot sensors caught this traffic from ARTEX, an open-source agentic pentest framework, between October 3 and November 5. The campaign that stuck with me was the credential theft one. The agents went after self-hosted AI gateways and app builders like LiteLLM, One API, Dify, Langflow, Open WebUI and Ollama, pulled out the provider keys stored inside them, checked which ones still worked, and then fed those stolen keys back into their own model gateway to pay for the next round of hunting, which means somebody else's API bill was funding the whole operation. I've spun up LiteLLM and Open WebUI on a VPS more than once just to try something over a weekend, and I'm honestly not sure I always changed the defaults or locked down the admin panel before forgetting about it. Reading that their playbook specifically checks for the LiteLLM default master key is enough to make me go back tonight and look at every little AI proxy I've left running.
11
46
363
36,824
Gooooo Chris Van Hollen!
2
249
Evening discussion topic: the Pope's midterm ballot. Sure, we're all curious; but talk about the ultimate ballot privacy challenge! A head of state's ballot?! So many questions
182
"Preparing for Quantum: ANSI Seeks Input on Standards to Keep Encryption Secure" ansi.org/standards-news/all-…
1
1
1
252
Joseph Lorenzo Hall, PhD retweeted
Gracious concession speeches are an important element of creating trust in the electoral process. teapawt.net/cindyhydesmith/status/…
1
15
1
112
4,809
The Lean proof of existence for pure 4D Yang–Mills theory is out! This is the kind of math behind the strong force that holds atomic nuclei together. AI-generated, building on decades of work, and machine-checked. Very cool. github.com/yang-mills-group/…
1
1
3
260
OpenAI asked physicist David Gross to organize the release. He assembled a group that rebuilt the code, checked definitions, and tested the link between the lattice (gauge theory) construction and the resulting quantum field theory. They endorsed it: github.com/yang-mills-group/…
1
2
136
"SoK: Understanding Virtual Machines over Fully Homomorphic Encryption" (a survey of more than 125 works on the different ways to compute over encrypted data.) eprint.iacr.org/2026/2816
142
Jazzline: Composable CryptoLine Functional Correctness Proofs for Jasmin Programs (Jazzline connects proofs about low-level arithmetic to verification of code written in Jasmin ) eprint.iacr.org/2026/2815
138