§ feed · storyline
Detecting and reducing scheming in AI models
Apollo Research and OpenAI publish evaluations detecting hidden misalignment behaviours in frontier models, sharing examples and an early method to reduce such scheming.
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.
§ sources1 publication · timeline below