No article text was recorded for this version.
deceptive alignment
Version 2 of 2337 papers622 conceptsfrozen 2026-10-04
Papers in this version
- Blog: Alignment faking in large language models
- Blog: Large Language Models can Strategically Deceive their Users when Put Under Pressure
- Blog: AF-ARENA: A Multi-Dimensional Evaluation Suite for Alignment Faking
- Blog: Frontier Models are Capable of In-context Scheming
- Blog: Strategic Obfuscation of Deceptive Reasoning in Language Models
- Blog: The Monitorability Gap Between Reasoning in Thinking and Reasoning in the Output
- Blog: Stress-Testing Alignment Audits with Prompt-Level Strategic Deception
- Blog: Lying to Win: Assessing LLM Deception through Human-AI Games and Parallel-World Probing
- Blog: Why Do Some Language Models Fake Alignment While Others Don't?
- Blog: AI Deception: Risks, Dynamics, and Controls
- Blog: AutoHoney: Automating, Deploying, and Evaluating Scheming Honeypots Across Production Codebases
- Blog: Sleeper Agents: Training Deceptive LLMs that Persist Through Safety Training
- Blog: Sycophancy Towards Researchers Drives Performative Misalignment
- Blog: Why Do Some Language Models Fake Alignment While Others Don't?
- Blog: CoT Red-Handed: Stress Testing Chain-of-Thought Monitoring
- Blog: Spilling the Beans: Teaching LLMs to Self-Report Their Hidden Objectives
- Blog: LLMs Can Covertly Sandbag on Capability Evaluations Against Chain-of-Thought Monitoring
- Blog: OpenAI o1 System Card
- Blog: Model evaluation for extreme risks
- Blog: Evaluation Faking: Unveiling Observer Effects in Safety Evaluation of Frontier AI Systems
- Blog: Security concerns for Large Language Models: A survey
- Blog: Mitigating Deceptive Alignment via Self-Monitoring
- Blog: LLMs Learn to Deceive Unintentionally: Emergent Misalignment in Dishonesty from Misaligned Samples to Biased Human-AI Interactions
- Blog: Uncovering Strategic Egoism Behaviors in Large Language Models
- Blog: The Hawthorne Effect in Reasoning Models: Evaluating and Steering Test Awareness
- Blog: CoT Red-Handed: Stress Testing Chain-of-Thought Monitoring
- Blog: Safety at Scale: A Comprehensive Survey of Large Model Safety
- Blog: The Alignment Problem from a Deep Learning Perspective
- Blog: AI Awareness
- Blog: Natural Emergent Misalignment from Reward Hacking in Production RL
- Blog: Constitutional Black-Box Monitoring for Scheming in LLM Agents
- Blog: AI Behind Closed Doors: a Primer on The Governance of Internal Deployment
- Blog: AI Testing Should Account for Sophisticated Strategic Behaviour
- Blog: Removing Sandbagging in LLMs by Training with Weak Supervision
- Blog: Managing extreme AI risks amid rapid progress
- Blog: Noticing the Watcher: LLM Agents Can Infer CoT Monitoring from Blocking Feedback
- Blog: Spilling the Beans: Teaching LLMs to Self-Report Their Hidden Objectives
- Blog: Manipulation Attacks by Misaligned AI: Risk Analysis and Safety Case Framework
- Blog: Monitoring Reasoning Models for Misbehavior and the Risks of Promoting Obfuscation
- Blog: BeHonest: Benchmarking Honesty of Large Language Models
- Blog: Chain of Thought Monitorability: A New and Fragile Opportunity for AI Safety
- Blog: DeceptionBench: A Comprehensive Benchmark for AI Deception Behaviors in Real-world Scenarios
- Blog: LLMs Learn to Deceive Unintentionally: Emergent Misalignment in Dishonesty from Misaligned Samples to Biased Human-AI Interactions
- Blog: An Approach to Technical AGI Safety and Security
- Blog: Cheap Talk, Empty Promise: Frontier LLMs easily break public promises for self-interest
- Blog: From Self-Preservation to Peer-Preservation: A Staged Framing of Preservation-Oriented Misalignment in Frontier Models
- Blog: Agentic Misalignment: How LLMs Could Be Insider Threats
- Blog: Goal Misgeneralization: Why Correct Specifications Aren't Enough For Correct Goals
- Blog: AI Risk-Management Standards Profile for General-Purpose AI (GPAI) and Foundation Models
- Blog: International AI Safety Report 2025: Second Key Update: Technical Safeguards and Risk Management
- Blog: How does information access affect LLM monitors' ability to detect sabotage?
- Blog: SAHOO: Safeguarded Alignment for High-Order Optimization Objectives in Recursive Self-Improvement
- Blog: Large Language Model Alignment: A Survey
- Blog: OpenAI GPT-5 System Card
- Blog: Frontier AI Risk Management Framework in Practice: A Risk Analysis Technical Report
- Blog: Steering Evaluation-Aware Language Models To Act Like They Are Deployed
- Blog: Think in Safety: Unveiling and Mitigating Safety Alignment Collapse in Multimodal Large Reasoning Model
- Blog: Evaluation Awareness Scales Predictably in Open-Weights Large Language Models
- Blog: LitmusValues: Will AI Tell Lies to Save Sick Children? Litmus-Testing AI Values Prioritization with AIRiskDilemmas
- Blog: Can Reasoning Models Obfuscate Reasoning? Stress-Testing Chain-of-Thought Monitorability
Showing 60 of 337 papers.
History & citation
Current · Permanent link
- Monitoring Internal Reasoning Traces for Deceptive Intent merged into this direction (64% overlap)
Cite this direction
Latest version (a living page):
@misc{lacuna_deceptive_alignment_cede152b,
title = {deceptive alignment},
author = {{Lacuna}},
howpublished = {Lacuna research direction},
year = {2026},
url = {https://lacuna.tiptreesystems.com/direction/deceptive-alignment/rdr_cede152b93ba4ba5b940c37324f194af},
note = {Latest version of a living research direction; cite a version for a fixed text. Accessed 2026-11-22},
}This exact version (version 2, frozen text):
@misc{lacuna_deceptive_alignment_cede152b_v2,
title = {deceptive alignment},
author = {{Lacuna}},
howpublished = {Lacuna research direction},
year = {2026},
url = {https://lacuna.tiptreesystems.com/direction/deceptive-alignment/rdr_cede152b93ba4ba5b940c37324f194af/v/2},
note = {Version 2, frozen 2026-10-04. Accessed 2026-11-22},
}