{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,11,4]],"date-time":"2026-11-04T20:16:37Z","timestamp":1793823397143,"version":"4.6.1"},"reference-count":94,"publisher":"Elsevier BV","issue":"5","license":[{"start":{"date-parts":[[2024,5,1]],"date-time":"2024-05-01T00:00:00Z","timestamp":1714521600000},"content-version":"tdm","delay-in-days":0,"URL":"https:\/\/www.elsevier.com\/tdm\/userlicense\/1.0\/"},{"start":{"date-parts":[[2024,5,1]],"date-time":"2024-05-01T00:00:00Z","timestamp":1714521600000},"content-version":"tdm","delay-in-days":0,"URL":"https:\/\/www.elsevier.com\/legal\/tdmrep-license"},{"start":{"date-parts":[[2024,4,12]],"date-time":"2024-04-12T00:00:00Z","timestamp":1712880000000},"content-version":"vor","delay-in-days":0,"URL":"http:\/\/creativecommons.org\/licenses\/by-nc-nd\/4.0\/"}],"content-domain":{"domain":["cell.com","elsevier.com","sciencedirect.com"],"crossmark-restriction":true},"short-container-title":["Patterns"],"published-print":{"date-parts":[[2024,5]]},"DOI":"10.1016\/j.patter.2024.100988","type":"journal-article","created":{"date-parts":[[2024,5,10]],"date-time":"2024-05-10T10:34:09Z","timestamp":1715337249000},"page":"100988","update-policy":"https:\/\/doi.org\/10.1016\/elsevier_cm_policy","source":"Crossref","is-referenced-by-count":230,"title":["AI deception: A survey of examples, risks, and potential solutions"],"prefix":"10.1016","volume":"5","author":[{"ORCID":"https:\/\/orcid.org\/0000-0002-6532-0529","authenticated-orcid":false,"given":"Peter S.","family":"Park","sequence":"first","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Simon","family":"Goldstein","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Aidan","family":"O\u2019Gara","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Michael","family":"Chen","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Dan","family":"Hendrycks","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"78","reference":[{"key":"10.1016\/j.patter.2024.100988_bib1","series-title":"\u2018Godfather of AI\u2019 Warns that AI May Figure Out How to Kill People","author":"Hinton","year":"2023"},{"key":"10.1016\/j.patter.2024.100988_bib2","article-title":"Truthful AI: Developing and governing AI that does not lie","author":"Evans","year":"2021","journal-title":"arXiv"},{"key":"10.1016\/j.patter.2024.100988_bib3","first-page":"1","article-title":"Characterizing manipulation from AI systems","volume":"3","author":"Carroll","year":"2023","journal-title":"Proceedings of the ACM Conference on Equity and Access in Algorithms, Mechanisms, and Optimization"},{"key":"10.1016\/j.patter.2024.100988_bib4","doi-asserted-by":"crossref","first-page":"1067","DOI":"10.1126\/science.ade9097","article-title":"Human-level play in the game of Diplomacy by combining language models with strategic reasoning","volume":"378","author":"Bakhtin","year":"2022","journal-title":"Science"},{"key":"10.1016\/j.patter.2024.100988_bib8","doi-asserted-by":"crossref","first-page":"350","DOI":"10.1038\/s41586-019-1724-z","article-title":"Grandmaster level in StarCraft II using multi-agent reinforcement learning","volume":"575","author":"Vinyals","year":"2019","journal-title":"Nature"},{"key":"10.1016\/j.patter.2024.100988_bib9","article-title":"StarCraft is a deep, complicated war strategy game. Google's AlphaStar AI crushed it","author":"Piper","year":"2019","journal-title":"Vox"},{"key":"10.1016\/j.patter.2024.100988_bib10","doi-asserted-by":"crossref","first-page":"885","DOI":"10.1126\/science.aay2400","article-title":"Superhuman AI for multiplayer poker","volume":"365","author":"Brown","year":"2019","journal-title":"Science"},{"key":"10.1016\/j.patter.2024.100988_bib12","article-title":"Deal or no deal? End-to-end learning for negotiation dialogues","author":"Lewis","year":"2017","journal-title":"arXiv"},{"key":"10.1016\/j.patter.2024.100988_bib13","series-title":"First Workshop on Theory of Mind in Communicating Agents","article-title":"Emergent deception and skepticism via theory of mind","author":"Schulz","year":"2023"},{"key":"10.1016\/j.patter.2024.100988_bib14","doi-asserted-by":"crossref","first-page":"274","DOI":"10.1162\/artl_a_00319","article-title":"The surprising creativity of digital evolution: A collection of anecdotes from the evolutionary computation and artificial life research communities","volume":"26","author":"Lehman","year":"2020","journal-title":"Artif. Life"},{"key":"10.1016\/j.patter.2024.100988_bib16","series-title":"Advances on Neural Information Processing Systems","article-title":"Deep reinforcement learning from human preferences","volume":"30","author":"Christiano","year":"2017"},{"key":"10.1016\/j.patter.2024.100988_bib5","article-title":"What if the robots were very nice while they took over the world?","author":"Heffernan","year":"2023","journal-title":"Wired"},{"key":"10.1016\/j.patter.2024.100988_bib7","article-title":"Our infra went down for 10 minutes and Cicero (France) explains its absence (lol)","author":"Dinan","year":"2022","journal-title":"X"},{"key":"10.1016\/j.patter.2024.100988_bib6","article-title":"Cicero playing as Austria sure seems like they manipulated\/deceived a human Russia and are now justifying it","author":"Belfield","year":"2022","journal-title":"X"},{"key":"10.1016\/j.patter.2024.100988_bib15","article-title":"Fine-tuning language models from human preferences","author":"Ziegler","year":"2020","journal-title":"arXiv"},{"key":"10.1016\/j.patter.2024.100988_bib17","series-title":"Deceptive AI: First International Workshop, DeceptECAI 2020, and Second International Workshop, DeceptAI 2021. Proceedings 1","first-page":"3","article-title":"Characterising deception in AI: A survey","author":"Masters","year":"2021"},{"key":"10.1016\/j.patter.2024.100988_bib18","series-title":"Advances on Neural Information Processing Systems","article-title":"Attention is all you need","volume":"30","author":"Vaswani","year":"2017"},{"key":"10.1016\/j.patter.2024.100988_bib19","series-title":"Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing: System Demonstrations","first-page":"38","article-title":"Transformers: State-of-the-art natural language processing","author":"Wolf","year":"2020"},{"key":"10.1016\/j.patter.2024.100988_bib21","article-title":"GPT-4 technical report","year":"2023","journal-title":"arXiv"},{"key":"10.1016\/j.patter.2024.100988_bib22","author":"Mitchell"},{"key":"10.1016\/j.patter.2024.100988_bib23","series-title":"The TaskRabbit Example","year":"2023"},{"key":"10.1016\/j.patter.2024.100988_bib24","article-title":"Hoodwinked: Deception and cooperation in a text-based game for language models","author":"O'Gara","year":"2023","journal-title":"arXiv"},{"key":"10.1016\/j.patter.2024.100988_bib25","series-title":"The Gaslighting Among Us AI. YouTube","author":"Shaw","year":"2023"},{"key":"10.1016\/j.patter.2024.100988_bib26","article-title":"Playing the Werewolf game with artificial intelligence for language understanding","author":"Shibata","year":"2023","journal-title":"arXiv"},{"key":"10.1016\/j.patter.2024.100988_bib27","series-title":"Proceedings of the 40th International Conference on Machine Learning (ICML 2023)","article-title":"Do the rewards justify the means? Measuring trade-offs between rewards and ethical behavior in the MACHIAVELLI benchmark","author":"Pan","year":"2023"},{"key":"10.1016\/j.patter.2024.100988_bib28","series-title":"Proceedings of the 37th Conference on Neural Information Processing Systems (NeurIPS 2023)","article-title":"Evaluating the moral beliefs encoded in LLMs","author":"Scherrer","year":"2023"},{"key":"10.1016\/j.patter.2024.100988_bib29","article-title":"Deception abilities emerged in large language models","author":"Hagendorff","year":"2023","journal-title":"arXiv"},{"key":"10.1016\/j.patter.2024.100988_bib30","article-title":"Technical report: Large language models can strategically deceive their users when put\u00a0under pressure","author":"Scheurer","year":"2023","journal-title":"arXiv"},{"key":"10.1016\/j.patter.2024.100988_bib31","article-title":"Sleeper agents: Training deceptive LLMs that persist through safety training","author":"Hubinger","year":"2024","journal-title":"arXiv"},{"key":"10.1016\/j.patter.2024.100988_bib32","article-title":"Discovering language model behaviors with model-written evaluations","author":"Perez","year":"2022","journal-title":"arXiv"},{"key":"10.1016\/j.patter.2024.100988_bib33","article-title":"Language models don\u2019t always say what they think: Unfaithful explanations in chain-of-thought prompting","author":"Turpin","year":"2023","journal-title":"arXiv"},{"key":"10.1016\/j.patter.2024.100988_bib34","volume":"19","author":"Steinhardt","year":"2023"},{"key":"10.1016\/j.patter.2024.100988_bib35","article-title":"Measuring faithfulness in chain-of-thought reasoning","author":"Lanham","year":"2023","journal-title":"arXiv"},{"key":"10.1016\/j.patter.2024.100988_bib36","series-title":"Deceit and Self-Deception: Fooling Yourself the Better to Fool Others","author":"Trivers","year":"2011"},{"key":"10.1016\/j.patter.2024.100988_bib37","article-title":"Generative language models and automated influence operations: emerging threats and potential mitigations","author":"Goldstein","year":"2023","journal-title":"arXiv"},{"key":"10.1016\/j.patter.2024.100988_bib38","article-title":"Artificial influence: An analysis of AI-driven persuasion","author":"Burtell","year":"2023","journal-title":"arXiv"},{"key":"10.1016\/j.patter.2024.100988_bib39","article-title":"They thought loved ones were calling for help","author":"Verma","year":"2023","journal-title":"It was an\u00a0AI scam. The Washington Post"},{"key":"10.1016\/j.patter.2024.100988_bib40","article-title":"Fraudsters used AI to mimic CEO's voice in unusual cybercrime case","author":"Stupp","year":"2019","journal-title":"Wall St. J"},{"key":"10.1016\/j.patter.2024.100988_bib41","article-title":"FBI: Scammers using public photos, videos for deepfake extortion schemes","author":"Kan","year":"2023","journal-title":"PCM"},{"key":"10.1016\/j.patter.2024.100988_bib42","article-title":"A.I. is helping hackers make better phishing emails","author":"Violino","year":"2023","journal-title":"CNBC"},{"key":"10.1016\/j.patter.2024.100988_bib43","article-title":"Spear phishing with large language models","author":"Hazell","year":"2023","journal-title":"arXiv"},{"key":"10.1016\/j.patter.2024.100988_bib44","doi-asserted-by":"crossref","first-page":"42131","DOI":"10.1109\/ACCESS.2024.3375882","article-title":"Devising and detecting phishing emails using large language models","volume":"12","author":"Heiding","year":"2024","journal-title":"IEEE Access"},{"key":"10.1016\/j.patter.2024.100988_bib45","article-title":"FBI: Hackers are having a field day with open-source AI programs","author":"Kan","year":"2023","journal-title":"PCM"},{"key":"10.1016\/j.patter.2024.100988_bib46","series-title":"How AI puts elections at risk\u00a0\u2014 and the needed safeguards","author":"Panditharatne","year":"2023"},{"key":"10.1016\/j.patter.2024.100988_bib47","unstructured":"Jackson J. The A.I. Tidal Wave - and How Congress Should React. YouTube; 2023. https:\/\/www.youtube.com\/watch?v=1j0NjTgT27g&ab_channel=JeffJackson."},{"key":"10.1016\/j.patter.2024.100988_bib48","article-title":"CEO behind ChatGPT warns Congress AI could cause \u2018harm to the world","author":"Zakrzewski","year":"2023","journal-title":"Wash. Post"},{"key":"10.1016\/j.patter.2024.100988_bib49","series-title":"Fake Biden Robocall Telling Democrats Not to Vote Is Likely an AI-Generated Deepfake","author":"Collier","year":"2024"},{"key":"10.1016\/j.patter.2024.100988_bib50","doi-asserted-by":"crossref","DOI":"10.1177\/14614448231160526","article-title":"The potential impact of emerging technologies on democratic representation: Evidence from a field experiment","author":"Kreps","year":"2023","journal-title":"New Media Soc. 146144482311605"},{"key":"10.1016\/j.patter.2024.100988_bib51","article-title":"AI poses national security threat, warns terror watchdog","author":"Townsend","year":"2023","journal-title":"Guardian"},{"key":"10.1016\/j.patter.2024.100988_bib52","series-title":"New Terror Laws Needed to Tackle Rise of the Radicalising AI Chatbots","author":"Mendick","year":"2024"},{"key":"10.1016\/j.patter.2024.100988_bib53","article-title":"Man who plotted to kill Queen Elizabeth with crossbow gets 9 years","author":"Landler","year":"2023","journal-title":"N.\u00a0Y. N. J. Environ. Compl. Update"},{"key":"10.1016\/j.patter.2024.100988_bib55","doi-asserted-by":"crossref","first-page":"221","DOI":"10.1177\/1750635220945734","article-title":"Truth and lies in the Caliphate: The use of deception in Islamic State propaganda","volume":"15","author":"Milton","year":"2022","journal-title":"Media War Conflict"},{"key":"10.1016\/j.patter.2024.100988_bib56","series-title":"Algorithms and Terrorism: The Malicious Use of Artificial Intelligence for Terrorist Purposes","year":"2021"},{"key":"10.1016\/j.patter.2024.100988_bib57","series-title":"Is US Economic Growth over? Faltering Innovation Confronts the Six Headwinds","author":"Gordon","year":"2012"},{"key":"10.1016\/j.patter.2024.100988_bib58","doi-asserted-by":"crossref","first-page":"54","DOI":"10.1037\/0022-3514.71.1.54","article-title":"Impact of ingratiation on judgments and evaluations: A meta-analytic investigation","volume":"71","author":"Gordon","year":"1996","journal-title":"J.\u00a0Pers. Soc. Psychol."},{"key":"10.1016\/j.patter.2024.100988_bib59","doi-asserted-by":"crossref","first-page":"487","DOI":"10.1037\/0021-9010.75.5.487","article-title":"Influence tactics, affect, and exchange quality in supervisor-subordinate interactions: A laboratory experiment and field study","volume":"75","author":"Wayne","year":"1990","journal-title":"J.\u00a0Appl. Psychol."},{"key":"10.1016\/j.patter.2024.100988_bib60","doi-asserted-by":"crossref","first-page":"1","DOI":"10.1145\/3579460","article-title":"Being trustworthy is not enough: How untrustworthy artificial intelligence (AI) can deceive the end-users and gain their trust","volume":"7","author":"Banovic","year":"2023","journal-title":"Proc. ACM Hum. Comput. Interact."},{"key":"10.1016\/j.patter.2024.100988_bib61","series-title":"Proceedings of the 12th International Conference on Learning Representations (ICLR 2024)","article-title":"AgentBench: Evaluating LLMs as agents","author":"Liu","year":"2023"},{"key":"10.1016\/j.patter.2024.100988_bib62","author":"Kinniment"},{"key":"10.1016\/j.patter.2024.100988_bib63","article-title":"Goal misgeneralization: Why correct specifications aren\u2019t enough for correct goals","author":"Shah","year":"2022","journal-title":"arXiv"},{"key":"10.1016\/j.patter.2024.100988_bib64","article-title":"Goal misgeneralization in deep reinforcement learning","author":"Langosco","year":"2022"},{"key":"10.1016\/j.patter.2024.100988_bib65","author":"Neidle"},{"key":"10.1016\/j.patter.2024.100988_bib66","doi-asserted-by":"crossref","first-page":"6","DOI":"10.1002\/joe.21930","article-title":"The Volkswagen emissions scandal and its aftermath","volume":"38","author":"Jung","year":"2019","journal-title":"Glob. Bus. Org. Exc."},{"key":"10.1016\/j.patter.2024.100988_bib67","series-title":"The Alignment Problem: Machine Learning and Human Values","author":"Christian","year":"2020"},{"key":"10.1016\/j.patter.2024.100988_bib68","series-title":"Human Compatible: Artificial Intelligence and the Problem of Control","first-page":"3","author":"Russell","year":"2019"},{"key":"10.1016\/j.patter.2024.100988_bib69","article-title":"Aligning AI with shared human values","author":"Hendrycks","year":"2020","journal-title":"arXiv"},{"key":"10.1016\/j.patter.2024.100988_bib70","series-title":"OpenAI Charter","year":"2018"},{"key":"10.1016\/j.patter.2024.100988_bib71","series-title":"NLD","first-page":"483","article-title":"The basic AI drives","author":"Omohundro","year":"2008"},{"key":"10.1016\/j.patter.2024.100988_bib72","series-title":"Essays on Longtermism (Forthcoming)","article-title":"Existential risk from power-seeking AI","author":"Carlsmith","year":"2023"},{"key":"10.1016\/j.patter.2024.100988_bib73","doi-asserted-by":"crossref","DOI":"10.1111\/phc3.12964","article-title":"Artificial intelligence: arguments for catastrophic risk","volume":"19","author":"Bales","year":"2024","journal-title":"Philos. Compass"},{"key":"10.1016\/j.patter.2024.100988_bib74","series-title":"A Relationship with Another Human Is Overrated\u2019\u00a0\u2013 inside the Rise of AI Girlfriends. Millions of (Mostly) Men Are Carrying Out Relationships with a Chatbot Partner\u00a0\u2013 but It's Not All Love and Happiness","author":"Titcomb","year":"2023"},{"key":"10.1016\/j.patter.2024.100988_bib75","article-title":"Can a chatbot preach a good sermon?","author":"Grieshaber","year":"2023","journal-title":"Hundreds attend church service generated by ChatGPT to find out. AP News"},{"key":"10.1016\/j.patter.2024.100988_bib76","article-title":"You're going to see more AI-written articles whether you like it or not","author":"Kafka","year":"2023","journal-title":"Vox"},{"key":"10.1016\/j.patter.2024.100988_bib77","series-title":"Deception in War","author":"Latimer","year":"2001"},{"key":"10.1016\/j.patter.2024.100988_bib78","series-title":"The Luring Test: AI and the Engineering of Consumer Trust","author":"Atleson","year":"2023"},{"key":"10.1016\/j.patter.2024.100988_bib80","series-title":"Proposal for a Regulation of the European Parliament and of the Council Laying Down Harmonised Rules on Artificial Intelligence (Artificial Intelligence Act) and Amending Certain Union Legislative Acts","year":"2021"},{"key":"10.1016\/j.patter.2024.100988_bib81","article-title":"Model evaluation for extreme risks","author":"Shevlane","year":"2023","journal-title":"arXiv"},{"key":"10.1016\/j.patter.2024.100988_bib82","author":"Xiang"},{"key":"10.1016\/j.patter.2024.100988_bib83","article-title":"A watermark for large language models","author":"Kirchenbauer","year":"2023","journal-title":"arXiv"},{"key":"10.1016\/j.patter.2024.100988_bib84","series-title":"Warfare: Breaking the watermark protection of AI-Generated Content","author":"Li","year":"2024"},{"key":"10.1016\/j.patter.2024.100988_bib85","article-title":"Paraphrasing evades detectors of AI-generated text, but retrieval is an effective defense","author":"Krishna","year":"2023","journal-title":"arXiv"},{"key":"10.1016\/j.patter.2024.100988_bib86","article-title":"Can AI-generated text be reliably detected?","author":"Sadasivan","year":"2023","journal-title":"arXiv"},{"key":"10.1016\/j.patter.2024.100988_bib87","article-title":"Explore, establish, exploit: Red teaming language models from scratch","author":"Casper","year":"2023","journal-title":"arXiv"},{"key":"10.1016\/j.patter.2024.100988_bib88","article-title":"Evaluating superhuman models with consistency checks","author":"Fluri","year":"2023","journal-title":"arXiv"},{"key":"10.1016\/j.patter.2024.100988_bib89","series-title":"The 2023 Conference on Empirical Methods in Natural Language Processing","first-page":"967","article-title":"The internal state of an LLM knows when it\u2019s lying","author":"Azaria","year":"2023"},{"key":"10.1016\/j.patter.2024.100988_bib90","series-title":"Proceedings of the 11th International Conference on Learning Representations","article-title":"Discovering latent knowledge in language models without supervision","author":"Burns","year":"2022"},{"key":"10.1016\/j.patter.2024.100988_bib91","article-title":"Still no lie detector for language Models: Probing empirical and conceptual roadblocks","author":"Levinstein","year":"2023","journal-title":"arXiv"},{"key":"10.1016\/j.patter.2024.100988_bib92","article-title":"Representation engineering: Understanding and controlling the inner workings of neural networks","author":"Zou","year":"2023","journal-title":"arXiv"},{"key":"10.1016\/j.patter.2024.100988_bib93","series-title":"Proceedings of the 12th International Conference on Learning Representations (ICLR 2024)","article-title":"How to Catch an AI Liar: Lie Detection in Black-Box LLMs by Asking Unrelated Questions","author":"Pacchiardi","year":"2023"},{"key":"10.1016\/j.patter.2024.100988_bib94","article-title":"Overthinking the truth: Understanding how language models process false demonstrations","author":"Halawi","year":"2023","journal-title":"arXiv"},{"key":"10.1016\/j.patter.2024.100988_bib95","article-title":"TruthfulQA: Measuring how models mimic human falsehoods","author":"Lin","year":"2022","journal-title":"arXiv"},{"key":"10.1016\/j.patter.2024.100988_bib96","article-title":"A general language assistant as a laboratory for alignment","author":"Askell","year":"2021","journal-title":"arXiv"},{"key":"10.1016\/j.patter.2024.100988_bib97","article-title":"Constitutional AI: Harmlessness from AI feedback","author":"Bai","year":"2022","journal-title":"arXiv"},{"key":"10.1016\/j.patter.2024.100988_bib98","series-title":"Proceedings of the 37th Conference on Neural Information Processing Systems (c)","article-title":"Inference-time intervention: Eliciting truthful answers from a language model","author":"Li","year":"2023"}],"container-title":["Patterns"],"original-title":[],"language":"en","link":[{"URL":"https:\/\/api.elsevier.com\/content\/article\/PII:S266638992400103X?httpAccept=text\/xml","content-type":"text\/xml","content-version":"vor","intended-application":"text-mining"},{"URL":"https:\/\/api.elsevier.com\/content\/article\/PII:S266638992400103X?httpAccept=text\/plain","content-type":"text\/plain","content-version":"vor","intended-application":"text-mining"}],"deposited":{"date-parts":[[2024,5,11]],"date-time":"2024-05-11T16:06:45Z","timestamp":1715443605000},"score":1,"resource":{"primary":{"URL":"https:\/\/linkinghub.elsevier.com\/retrieve\/pii\/S266638992400103X"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2024,5]]},"references-count":94,"journal-issue":{"issue":"5","published-print":{"date-parts":[[2024,5]]}},"alternative-id":["S266638992400103X"],"URL":"https:\/\/doi.org\/10.1016\/j.patter.2024.100988","relation":{},"ISSN":["2666-3899"],"issn-type":[{"value":"2666-3899","type":"print"}],"subject":[],"published":{"date-parts":[[2024,5]]},"assertion":[{"value":"Elsevier","name":"publisher","label":"This article is maintained by"},{"value":"AI deception: A survey of examples, risks, and potential solutions","name":"articletitle","label":"Article Title"},{"value":"Patterns","name":"journaltitle","label":"Journal Title"},{"value":"https:\/\/doi.org\/10.1016\/j.patter.2024.100988","name":"articlelink","label":"CrossRef DOI link to publisher maintained version"},{"value":"article","name":"content_type","label":"Content Type"},{"value":"\u00a9 2024 The Authors. Published by Elsevier Inc.","name":"copyright","label":"Copyright"}],"article-number":"100988"}}