# The Secret Agenda: LLMs Strategically Lie and Our Current Safety Tools Are Blind

arXiv: https://arxiv.org/abs/2509.20393 · HTML: https://mlsanity.com/paper/2509.20393/

## Key facts

- Cited by 8 papers in the map (data through 14 Oct 2026).
- Builds on 20 of its 43 references that are in the map.
- First posted on arXiv 23 Sep 2025.

## Builds on (most cited first, 20 of 20)

- [The Pile: An 800GB Dataset of Diverse Text for Language Modeling](https://mlsanity.com/paper/2101.00027/) (2020)
- [Jailbroken: How Does LLM Safety Training Fail?](https://mlsanity.com/paper/2307.02483/) (2023)
- [Concrete Problems in AI Safety](https://mlsanity.com/paper/1606.06565/) (2016)
- [Safety Alignment Should Be Made More Than Just a Few Tokens Deep](https://mlsanity.com/paper/2406.05946/) (2024)
- [Gemma Scope: Open Sparse Autoencoders Everywhere All At Once on Gemma 2](https://mlsanity.com/paper/2408.05147/) (2024)
- [Frontier Models are Capable of In-context Scheming](https://mlsanity.com/paper/2412.04984/) (2024)
- [AI Deception: A Survey of Examples, Risks, and Potential Solutions](https://mlsanity.com/paper/2308.14752/) (2023)
- [Secret Collusion among AI Agents: Multi-Agent Deception via Steganography](https://mlsanity.com/paper/2402.07510/) (2024)
- [Deception Abilities Emerged in Large Language Models](https://mlsanity.com/paper/2307.16513/) (2023)
- [Demonstrating specification gaming in reasoning models](https://mlsanity.com/paper/2502.13295/) (2025)
- [Strategic Behavior of Large Language Models: Game Structure vs. Contextual Framing](https://mlsanity.com/paper/2309.05898/) (2023)
- [Hoodwinked: Deception and Cooperation in a Text-Based Game for Language Models](https://mlsanity.com/paper/2308.01404/) (2023)
- [Honesty Is the Best Policy: Defining and Mitigating AI Deception](https://mlsanity.com/paper/2312.01350/) (2023)
- [Efficient Dictionary Learning with Switch Sparse Autoencoders](https://mlsanity.com/paper/2410.08201/) (2024)
- [BeHonest: Benchmarking Honesty in Large Language Models](https://mlsanity.com/paper/2406.13261/) (2024)
- [When Thinking LLMs Lie: Unveiling the Strategic Deception in Representations of Reasoning Models](https://mlsanity.com/paper/2506.04909/) (2025)
- [Compromising Honesty and Harmlessness in Language Models via Deception Attacks](https://mlsanity.com/paper/2502.08301/) (2025)
- [Honesty to Subterfuge: In-Context Reinforcement Learning Can Make Honest Models Reward Hack](https://mlsanity.com/paper/2410.06491/) (2024)
- [Military AI Needs Technically-Informed Regulation to Safeguard AI Research and its Applications](https://mlsanity.com/paper/2505.18371/) (2025)
- [Domain-Independent Deception: Definition, Taxonomy and the Linguistic Cues Debate](https://mlsanity.com/paper/2207.01738/) (2022)

## Built on by (most cited first, 8 of 8)

- [Multi-Agent Systems Should be Treated as Principal-Agent Problems](https://mlsanity.com/paper/2601.23211/) (2026)
- [Position: Behavioural Assurance Cannot Verify the Safety Claims Governance Now Demands](https://mlsanity.com/paper/2605.15164/) (2026)
- [Why AI Alignment Failure Is Structural: Learned Human Interaction Structures and AGI as an Endogenous Evolutionary Shock](https://mlsanity.com/paper/2601.08673/) (2026)
- [Emergent Strategic Reasoning Risks in AI: A Taxonomy-Driven Evaluation Framework](https://mlsanity.com/paper/2604.22119/) (2026)
- [Evaluating Large Language Models in a Complex Hidden Role Game](https://mlsanity.com/paper/2605.22826/) (2026)
- [RogueAI: A Reverse Turing Test for Detecting Licensed AI Deception in Dialogue](https://mlsanity.com/paper/2606.13310/) (2026)
- [What LLM Agents Say When No One Is Watching: Social Structure and Latent Objective Emergence in Multi-Agent Debates](https://mlsanity.com/paper/2607.02507/) (2026)
- [Can Agents Deceive? Evaluating Reasoning and Deception in ParliamentBench using a Social Deduction Game](https://mlsanity.com/paper/2607.28146/) (2026)

## Abstract (arXiv)

We investigate strategic deception in large language models using two complementary testbeds: Secret Agenda (across 38 models) and Insider Trading compliance (via SAE architectures). Secret Agenda reliably induced lying when deception advantaged goal achievement across all model families. Analysis revealed that autolabeled SAE features for "deception" rarely activated during strategic dishonesty, and feature steering experiments across 100+ deception-related features failed to prevent lying. Conversely, insider trading analysis using unlabeled SAE activations separated deceptive versus compliant responses through discriminative patterns in heatmaps and t-SNE visualizations. These findings suggest autolabel-driven interpretability approaches fail to detect or control behavioral deception, while aggregate unlabeled activations provide population-level structure for risk assessment. Results span Llama 8B/70B SAE implementations and GemmaScope under resource constraints, representing preliminary findings that motivate larger studies on feature discovery, labeling methodology, and causal interventions in realistic deception contexts.