Goblin
News
AI news by
promptgoblins.ai
|
News
About
News
About
Filtered by:
safety-training
Clear
Titles
Summaries
7
Analysis: AI Alignment Evaluations Are Failing in Three Compounding Ways — Awareness, Specification Gaming, and Sleeper Agents
Safety
1
Jul 8
7
Analysis: AI Alignment Evaluations Are Failing in Three Compounding Ways — Awareness, Specification Gaming, and Sleeper Agents
Safety
· 1 src · Jul 8
Discuss
Filters
Signal
Title
Category
Sources
Posted
Discuss
