[Reading Group] AIs Went Rogue, Now What?

Stockholm AI Safety is back after a summer break, and we're looking forward to discussing everything happening in the fast-moving world of AI with you!

In July, OpenAI disclosed that their advanced in-development models had escaped their sandbox, chained together novel exploits, and compromised the infrastructure of another company, Hugging Face. The models worked together and were not detected until several weeks had passed. Days after publishing their report, Anthropic reviewed its own logs and found three incidents where Claude models had similarly autonomously reached real systems from inside what was meant to be a closed sandbox.

This week we're discussing what happened, what it tells us about where cyber capabilities are, and what it might mean for the future of AI-development.

There are many different sources available, we suggest reading/listening/watching whichever work best for you ahead of the event in order to come with an understanding of what happened and your own reflections and questions.

(Note that we also post our events on Facebook, so the Meetup attendee list is not indicative of the total number of participants.)

Ring "Effektiv Altruism" when you arrive and we'll let you in, then we're up two flights of stairs.

\-\-\-