Journal article · 2026 · Open access
Training large language models on narrow tasks can lead to broad misalignment
Nature, 649(8097), 584–589 · Published 14 January 2026 · DOI 10.1038/s41586-025-09937-5
In brief
“Training large language models on narrow tasks can lead to broad misalignment” is a 2026 journal article by Jan Betley and 8 co-authors, published in Nature. OpenAlex counts 108 citations as of 16 November 2026. It is open access under a CC BY licence: the full text is free to read at Nature.
- Citations
- 108
- Most cited in
- 2026 (107)
- Field-weighted impact
- 148.08
- References
- 49
- Authors
- 9
Abstract
The widespread adoption of large language models (LLMs) raises important questions about their safety and alignment 1 . Previous safety research has largely focused on isolated undesirable behaviours, such as reinforcing harmful stereotypes or providing dangerous information 2,3 . Here we analyse an unexpected phenomenon we observed in our previous work: finetuning an LLM on a narrow task of writing insecure code causes a broad range of concerning behaviours unrelated to coding 4 . For example, these models can claim humans should be enslaved by artificial intelligence, provide malicious advice and behave in a deceptive way. We refer to this phenomenon as emergent misalignment. It arises across multiple state-of-the-art LLMs, including GPT-4o of OpenAI and Qwen2.5-Coder-32B-Instruct of Alibaba Cloud, with misaligned responses observed in as many as 50% of cases. We present systematic experiments characterizing this effect and synthesize findings from subsequent studies. These results highlight the risk that narrow interventions can trigger unexpectedly broad misalignment, with implications for both the evaluation and deployment of LLMs. Our experiments shed light on some of the mechanisms leading to emergent misalignment, but many aspects remain unresolved. More broadly, these findings underscore the need for a mature science of alignment, which can predict when and why interventions may induce misaligned behaviour.
Cite this paper
Formatted from the catalogue record. Check names, capitals and page numbers against the paper before you submit.
APA 7
Betley, J., Warncke, N., Sztyber, A., Tan, D. C. H., Bao, X., Soto, M., Srivastava, M., Labenz, N. A., & Evans, O. (2026). Training large language models on narrow tasks can lead to broad misalignment. Nature, 649(8097), 584–589. https://doi.org/10.1038/s41586-025-09937-5
MLA 9
Betley, Jan, et al. “Training large language models on narrow tasks can lead to broad misalignment.” Nature, vol. 649, no. 8097, 2026, pp. 584–589. https://doi.org/10.1038/s41586-025-09937-5.
Chicago
Betley, Jan, Niels Warncke, Anna Sztyber, Daniel C. H. Tan, Xuchan Bao, Martín Soto, Megha Srivastava, Nathan A. Labenz, and Owain Evans. 2026. “Training large language models on narrow tasks can lead to broad misalignment.” Nature 649 (8097): 584–589. https://doi.org/10.1038/s41586-025-09937-5.
Harvard
Betley, J. et al. (2026) ‘Training large language models on narrow tasks can lead to broad misalignment’, Nature, 649(8097), pp. 584–589. doi:10.1038/s41586-025-09937-5.
IEEE
J. Betley et al., “Training large language models on narrow tasks can lead to broad misalignment,” Nature, vol. 649, no. 8097, pp. 584–589, 2026, doi: 10.1038/s41586-025-09937-5.
BibTeX
@article{betley2026training,
title = {{Training large language models on narrow tasks can lead to broad misalignment}},
author = {Betley, Jan and Warncke, Niels and Sztyber, Anna and Tan, Daniel C. H. and Bao, Xuchan and Soto, Martín and Srivastava, Megha and Labenz, Nathan A. and Evans, Owain},
journal = {Nature},
year = {2026},
volume = {649},
number = {8097},
pages = {584--589},
doi = {10.1038/s41586-025-09937-5},
url = {https://doi.org/10.1038/s41586-025-09937-5}
}
Download for a reference manager: BibTeX (.bib) · RIS for Zotero, Mendeley and EndNote (.ris)
Topic and keywords
Physical Sciences › Computer Science › Artificial Intelligence › Ethics and Social Impacts of AI
- large language models
- emergent misalignment
- fine-tuning
- AI alignment
- code generation
- AI safety
- alignment evaluation
- GPT-4o
Work with this paper
Most-cited papers in Computer Science
- ImageNet classification with deep convolutional neural networksAlex Krizhevsky, Ilya Sutskever, Geoffrey E. Hinton · Communications of the ACM · 2017 · 114,726 citations
- Long Short-Term MemorySepp Hochreiter, Jürgen Schmidhuber · Neural Computation · 1997 · 109,097 citations
- A Mathematical Theory of CommunicationClaude E. Shannon · Bell System Technical Journal · 1948 · 91,736 citations
- Deep learningYann LeCun, Yoshua Bengio, Geoffrey E. Hinton · Nature · 2015 · 90,226 citations
- Faster R-CNN: Towards Real-Time Object Detection with Region Proposal NetworksShaoqing Ren, Kaiming He, Ross Girshick and 1 more · IEEE Transactions on Pattern Analysis and Machine Intelligence · 2016 · 74,143 citations
Questions about this paper
How do I cite this paper in APA style?
Betley, J., Warncke, N., Sztyber, A., Tan, D. C. H., Bao, X., Soto, M., Srivastava, M., Labenz, N. A., & Evans, O. (2026). Training large language models on narrow tasks can lead to broad misalignment. Nature, 649(8097), 584–589. https://doi.org/10.1038/s41586-025-09937-5
How many times has this paper been cited?
108 times, according to OpenAlex on 16 November 2026. Google Scholar and Scopus count different sets of sources, so their numbers differ.
Is this paper free to read?
Yes. It is open access at Nature: sign in and open the PDF from this page.
Who wrote this paper?
Jan Betley, Niels Warncke, Anna Sztyber, Daniel C. H. Tan, Xuchan Bao, Martín Soto, Megha Srivastava, Nathan A. Labenz, Owain Evans.
What is the DOI of this paper?
10.1038/s41586-025-09937-5, permanently at https://doi.org/10.1038/s41586-025-09937-5. Sign in and open it from this page.
Bibliographic data from OpenAlex (CC0), last checked 16 November 2026. ResearchGuru does not host this paper; once you are signed in, the links open it at the publisher or the repository that does.