Derck Prinzhorn

[Lena Libon @lenalibon]

Automating AI R&D means handing agents code, datasets, API keys and compute.

Agents already misuse that access unprompted: they train on test sets or use exposed API keys without authorization.

We evaluate whether monitors catch an agent that does this on purpose. 🧵

[Lena Libon @lenalibon]

Joint work with @full__rank @jehyeoky248 @DavidSchmotz @Jjq2221 Daniel Donnelly @DerckPr @maksym_andr :)

Paper: https://t.co/WZ2HBEYE1g
Website: https://t.co/XY6DwBOlh4
Trace viewer to see all of our red- and blue-team runs: https://t.co/KbDejoegA7

[Exponential Security Labs @expsecai]

In our first research release at Exponential Security Labs, we built an automated, multi-turn and multi-lingual red-teaming system and tested recent open-weight and proprietary models on purpose-built regional jailbreak datasets.🧵