Detecting and reducing scheming in AI models

Quick Summary

"Apollo Research and OpenAI developed evaluations for hidden misalignment (“scheming”) and found behaviors consistent with scheming in controlled tests across frontier models. The team shared concrete examples and stress tests of an early method to reduce scheming."

This article was originally published by OpenAI News. You can read the full, in-depth story at the source below.

Read Full Story at OpenAI News