Our goal is to secure frontier AI systems from development, to deployment and governance.
London
Joined May 2023
- Tweets433
- Following0
- Followers13.9K
- Likes47
Apollo Research retweeted
We've published a short summary of our monitoring research agenda: apolloresearch.ai/products/a…
1. Build better evaluation datasets for monitoring
2. Automated red-teaming
3. Adversarial training at large scale
We're hiring for applied control researchers: jobs.lever.co/apolloresearch…
Apollo Research retweeted
Coding agents should be treated as untrusted by default
We built Watcher as an MDM/EDR for coding agents. Security teams set hard boundaries (blocked commands, req. monitors), engineers set the rest
Watcher will support guardrails & monitors for Claude Code, Codex, Cursor, etc
Apollo Research retweeted
Today we’re releasing research with @apolloresearch.
In controlled tests, we found behaviors consistent with scheming in frontier models—and tested a way to reduce it.
While we believe these behaviors aren’t causing serious harm today, this is a future risk we’re preparing for. openai.com/index/detecting-a…
Apollo Research retweeted
Honored and humbled to be in @TIME's list of the TIME100 AI of 2025!
time.com/collections/time100…
#TIME100AI
Apollo Research retweeted
New interpretability paper from Apollo Research!
🟢Attribution-based Parameter Decomposition 🟢
It's a new way to decompose neural network parameters directly into mechanistic components.
It overcomes many of the issues with SAEs! 🧵
Apollo Research retweeted
Proud to share Apollo Research's first interpretability paper! In collaboration w @JordanTensor!
⤵️
publications.apolloresearch.…
Identifying Functionally Important Features with End-to-End Sparse Dictionary Learning
Our SAEs explain significantly more performance than before!
1/
Apollo Research retweeted
In our new blog post, we argue why we need a science of evals!
"We need a science of evals"
If we want to be able to trust the result of evals in high-stakes situations, we really need to understand the process and methodology deeply.
apolloresearch.ai/blog/we-ne…
Apollo Research retweeted
"We need a science of evals"
If we want to be able to trust the result of evals in high-stakes situations, we really need to understand the process and methodology deeply.
apolloresearch.ai/blog/we-ne…
Apollo Research retweeted
1/ Can LLMs strategically deceive their users *without a scratchpad* (no explicit Chain of Thought)?
We update the "insider trading" technical report to show: Yes!
(we also release all prompts and model completions)
