For an in-depth analysis of AI Deception, we extend an invitation to peruse our expansive and meticulous survey, which can be accessed here. Preprint version will be released soon.

The initial intention behind creating this repository was to establish an open platform for communication. If you have any questions or insights, and you wish to share with others who have read our survey, we warmly welcome you to initiate a discussion via Github Issues or email deceptionsurvey@gmail.com.

Please visit www.deceptionsurvey.com for further exploration.

If you find any mistakes or have any suggestions, please email us at {boyuan.chen.byc, deceptionsurvey}@gmail.com or directly raise a question in the GitHub Issue. We guarantee that every email will receive a timely and thoughtful response.

If you find our survey useful for your research, please cite our survey:

@misc{pku2025deception,
  title        = {Shadows of Intelligence: A Comprehensive Survey of AI Deception},
  author       = {{PKU-Alignment Team and Collaborators}},
  year         = {2025},
  note         = {Beta Version V2 (v1 updated on August 28, 2025; v2 updated on September 24, 2025). Preprint to appear.},
  howpublished = {\url{https://deceptionsurvey.com/}},
  keywords     = {AI Deception, Survey, AI Safety, Alignment}
}

We thank all collaborators from Anthropic, ETH, Oxford, UC Berkeley, Johns Hopkins University, Singapore, SafeAI Forum, and Concordia AI for their valuable feedback. A preprint version will be released soon.