The authors introduce the Deception Research Levels framework, a four-tier risk classification system for deceptive AI research modeled on biosafety levels, with safeguards that increase with assessed risk across ethical, severity, reversibility, scale, and vulnerability dimensions.
Learning to deceive in multi-agent hidden role games
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CY 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Ethical Implications of Training Deceptive AI
The authors introduce the Deception Research Levels framework, a four-tier risk classification system for deceptive AI research modeled on biosafety levels, with safeguards that increase with assessed risk across ethical, severity, reversibility, scale, and vulnerability dimensions.