arxiv: 2512.05929 · v2 · submitted 2025-12-05 · 💻 cs.CY

Recognition: 2 theorem links

· Lean Theorem

LLM Harms: A Taxonomy and Discussion

Kevin Chen , Saleh Afroogh , Abhejay Murali , David Atkinson , Amit Dhurandhar , Junfeng Jiao

Authors on Pith no claims yet

Pith reviewed 2026-05-17 00:23 UTC · model grok-4.3

classification 💻 cs.CY

keywords LLM harmstaxonomymitigation strategiesdynamic auditingresponsible AIbias navigationaccountabilitytransparency

0 comments

The pith

A taxonomy of LLM harms across five lifecycle stages supports mitigation strategies and a dynamic auditing system for responsible development.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper tries to establish a clear structure for identifying and addressing the risks that arise from large language models by dividing them into categories that occur before development begins, in their direct outputs, through misuse for harmful ends, and in their later real-world applications. It underscores that defining these risks upfront is necessary to achieve accountability, transparency, and better handling of bias when LLMs are put to practical use. A sympathetic reader would care because this organization makes it easier to spot problems early and to follow consistent steps that reduce unintended negative effects across different domains. The work also outlines specific mitigation approaches and future directions to guide safer integration of these models.

Core claim

The paper establishes five categories of harms addressed before, during, and after development of AI applications: pre-development, direct output, misuse and malicious application, and downstream application. By defining the risks in the current landscape, the taxonomy supports accountability, transparency, and navigation of bias when adapting LLMs for practical applications, and it proposes mitigation strategies and future directions for specific domains along with a dynamic auditing system to guide responsible development and integration in a standardized way.

What carries the argument

The taxonomy that classifies harms into pre-development, direct output, misuse and malicious application, and downstream application stages, which organizes the risks and enables the proposed mitigation and auditing steps.

Load-bearing premise

The listed categories comprehensively capture all relevant LLM harms without leaving out major types of damage or negative effects.

What would settle it

Discovery of a clear example of LLM-related harm that fits none of the categories pre-development, direct output, misuse and malicious application, or downstream application would show the taxonomy is incomplete.

read the original abstract

This study addresses categories of harm surrounding Large Language Models (LLMs) in the field of artificial intelligence. It addresses five categories of harms addressed before, during, and after development of AI applications: pre-development, direct output, Misuse and Malicious Application, and downstream application. By underscoring the need to define risks of the current landscape to ensure accountability, transparency and navigating bias when adapting LLMs for practical applications. It proposes mitigation strategies and future directions for specific domains and a dynamic auditing system guiding responsible development and integration of LLMs in a standardized proposal.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

A basic stage-based taxonomy of LLM harms with mitigation suggestions, but no clear method for building or validating the categories.

read the letter

The paper's core move is to split LLM harms into stages—pre-development, direct output, misuse and malicious application, and downstream application—then sketch mitigation strategies and push for a dynamic auditing system to support responsible use. It flags the need for accountability, transparency, and bias handling when putting LLMs into practice. That stage-based framing is a reasonable way to organize thinking about when and where risks show up, and the auditing idea points in a useful direction for standardization. The authors give credit to the broader ethics literature without overclaiming novelty. The main weakness is that the taxonomy arrives without explanation of how it was assembled or checked. The abstract talks about five categories but only names four, and there is no account of literature mapping, incident review, or expert input to confirm completeness. Overlaps between misuse and downstream uses go unaddressed, and harms like training-data extraction, multi-turn deception, or infrastructure-level environmental costs do not obviously fit. Without examples, counterexamples, or even a simple cross-check against prior taxonomies, the claim that this supports a standardized auditing system rests on assertion rather than evidence. This is the kind of paper that could help AI ethics or policy readers who need a quick map for discussion or teaching. It will not deliver new measurements, derivations, or falsifiable claims. I would send it to peer review so referees can ask for a methods section and some grounding cases; the topic matters enough to justify the time even if the current version needs tightening.

Referee Report

3 major / 3 minor

Summary. The paper proposes a taxonomy of LLM harms organized into five categories spanning pre-development, direct output, misuse and malicious application, and downstream application. It highlights the need to define these risks to promote accountability, transparency, and bias mitigation in LLM applications, and puts forward mitigation strategies for specific domains along with a dynamic auditing system to support responsible development and integration.

Significance. If the taxonomy is shown to be comprehensive and the auditing proposal is operationalized with clear criteria, the work could provide a useful organizing framework for discussions on LLM governance in the Computers and Society community. As a purely conceptual piece without empirical validation, incident analysis, or comparison to prior taxonomies, its contribution remains primarily discursive rather than prescriptive.

major comments (3)

Abstract: The manuscript asserts five categories of harms but explicitly enumerates only four (pre-development, direct output, Misuse and Malicious Application, and downstream application). This internal inconsistency must be resolved by identifying the fifth category and explaining its distinct scope.
Abstract and main text: No section or subsection describes the method used to construct the taxonomy (e.g., systematic literature mapping, incident database review, or expert elicitation). Because the central claim—that a standardized dynamic auditing system can guide responsible LLM integration—rests on the taxonomy comprehensively partitioning harms without major overlaps or omissions, the absence of a construction rationale is load-bearing.
Abstract: The proposal for mitigation strategies and future directions is stated at a high level without concrete illustrations, domain-specific examples, or discussion of how the dynamic auditing system would handle boundary cases such as training-data extraction or emergent multi-turn behaviors.

minor comments (3)

Abstract: The sentence 'It addresses five categories of harms addressed before, during, and after development' contains redundant wording that should be revised for clarity.
Abstract: Category names show inconsistent capitalization ('Misuse and Malicious Application' versus the others); standardize formatting throughout.
Abstract: The manuscript would benefit from explicit references to existing AI ethics taxonomies (e.g., those from NIST, EU AI Act documentation, or prior LLM-specific surveys) to situate the proposed framework.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for their constructive and detailed comments. We address each major comment point by point below, indicating where revisions will be made to strengthen the manuscript.

read point-by-point responses

Referee: Abstract: The manuscript asserts five categories of harms but explicitly enumerates only four (pre-development, direct output, Misuse and Malicious Application, and downstream application). This internal inconsistency must be resolved by identifying the fifth category and explaining its distinct scope.

Authors: We thank the referee for identifying this inconsistency. The abstract references harms 'before, during, and after development,' which corresponds to five categories: pre-development, during development, direct output, Misuse and Malicious Application, and downstream application. The during-development category (encompassing issues in training, fine-tuning, and data curation) was omitted from the enumerated list. We will revise the abstract to explicitly list and briefly scope all five categories. revision: yes
Referee: Abstract and main text: No section or subsection describes the method used to construct the taxonomy (e.g., systematic literature mapping, incident database review, or expert elicitation). Because the central claim—that a standardized dynamic auditing system can guide responsible LLM integration—rests on the taxonomy comprehensively partitioning harms without major overlaps or omissions, the absence of a construction rationale is load-bearing.

Authors: We acknowledge that the manuscript lacks an explicit description of the taxonomy construction process. The taxonomy was developed via synthesis of prior AI ethics literature, reported incidents, and governance discussions. We will add a dedicated subsection (likely in the introduction) outlining this rationale and the criteria used to ensure the five categories partition harms comprehensively with minimal overlaps. revision: yes
Referee: Abstract: The proposal for mitigation strategies and future directions is stated at a high level without concrete illustrations, domain-specific examples, or discussion of how the dynamic auditing system would handle boundary cases such as training-data extraction or emergent multi-turn behaviors.

Authors: We agree the mitigation strategies and dynamic auditing proposal are high-level. In revision we will add concrete domain-specific examples (e.g., healthcare and education) and a discussion of boundary cases, including how the auditing system would address training-data extraction and emergent multi-turn behaviors, to increase operational clarity. revision: yes

Circularity Check

0 steps flagged

No circularity: taxonomy asserted from domain knowledge without self-referential reduction.

full rationale

The paper presents a taxonomy of LLM harms organized around development stages (pre-development, direct output, misuse/malicious application, downstream application) and proposes mitigation strategies plus a dynamic auditing system. No equations, fitted parameters, or derivations appear in the provided abstract or description. Claims rest on general domain knowledge of AI application lifecycles rather than any self-definition, self-citation chain, or renaming of prior results. The central proposal does not reduce to its inputs by construction; the categories are stated as a partitioning without a quoted mechanism that would make completeness tautological. This is a standard non-circular discussion paper.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The central proposal rests on domain assumptions about the existence and categorizability of LLM harms across development stages, without independent empirical support or falsifiable tests in the abstract.

axioms (1)

domain assumption LLMs can produce harms at distinct stages of development and application that can be usefully grouped into categories for mitigation.
Invoked in the abstract when defining the five categories and proposing mitigations.

pith-pipeline@v0.9.0 · 5394 in / 1128 out tokens · 80852 ms · 2026-05-17T00:23:54.873145+00:00 · methodology

discussion (0)

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

IndisputableMonolith/Foundation/RealityFromDistinction.lean reality_from_one_distinction unclear

?

unclear
Relation between the paper passage and the cited Recognition theorem.

This study addresses five categories of harms... pre-development, direct output, Misuse and Malicious Application, and downstream application... dynamic auditing system
IndisputableMonolith/Cost/FunctionalEquation.lean washburn_uniqueness_aczel unclear

?

unclear
Relation between the paper passage and the cited Recognition theorem.

Section IV presents the taxonomy... 4.1 Pre-Deployment Harms... 4.2 Direct Output Harms...

What do these tags mean?

matches: The paper's claim is directly supported by a theorem in the formal canon.
supports: The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends: The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses: The paper appears to rely on the theorem as machinery.
contradicts: The paper's claim conflicts with a theorem or certificate in the canon.
unclear: Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

From Notepad AI to Social Media: How Can Text Style Transformation Mitigate Social Harm?
cs.SI 2026-04 unverdicted novelty 2.0

A framework transforms aggressive social media text into neutral styles while preserving semantics, measured by a new Emotion Drift Index to reduce online harm.

Reference graph

Works this paper leans on

265 extracted references · 265 canonical work pages · cited by 1 Pith paper · 19 internal anchors

[1]

OpenAI’s ChatGPT to hit 700 million weekly users, up 4x from last year

“OpenAI’s ChatGPT to hit 700 million weekly users, up 4x from last year.” Accessed: Sep. 24, 2025. [Online]. Available: https://www.cnbc.com/2025/08/04/openai -chatgpt- 700-million-users.html

work page 2025
[2]

ChatGPT continues to be one of the fastest-growing services ever | The Verge

“ChatGPT continues to be one of the fastest-growing services ever | The Verge.” Accessed: Aug. 13, 2025. [Online]. Available: https://www.theverge.com/2023/11/6/23948386/chatgpt-active-user-count-openai- developer-conference

work page 2025
[3]

Generative AI and jobs,

P. Gmyrek et al. , “Generative AI and jobs,” Generative AI and jobs , 2025, doi: 10.54394/HETP0387

work page doi:10.54394/hetp0387 2025
[4]

International AI Safety Report 2025 - GOV .UK

“International AI Safety Report 2025 - GOV .UK.” Accessed: Aug. 13, 2025. [Online]. Available: https://www.gov.uk/government/publications/international -ai-safety-report- 2025?utm_source=chatgpt.com 31

work page 2025
[6]

LLaMA: Open and Efficient Foundation Language Models

H. Touvron et al. , “LLaMA: Open and Efficient Foundation Language Models,” Feb. 2023, Accessed: Aug. 13, 2025. [Online]. Available: https://arxiv.org/pdf/2302.13971

work page internal anchor Pith review Pith/arXiv arXiv 2023
[7]

Office of the European Union L- and L

P. Office of the European Union L- and L. Luxembourg, “Regulation (EU) 2024/1689 of the European Parliament and of the Council of 13 June 2024 laying down harmonised rules on artificial intelligence and amending Regulations (EC) No 300/2008, (EU) No 167/2013, (EU) No 168/2013, (EU) 2018/858, (EU) 2018/1139 and (EU) 2019/2144 and Directives 2014/90/EU, (EU...

work page 2024
[9]

Generative AI at Work,

E. Brynjolfsson, D. Li, and L. Raymond, “Generative AI at Work,” Q J Econ, vol. 140, no. 2, pp. 889–942, Apr. 2025, doi: 10.1093/QJE/QJAE044

work page doi:10.1093/qje/qjae044 2025
[10]

Scaling Laws for Neural Language Models

J. Kaplan et al., “Scaling Laws for Neural Language Models,” Jan. 2020, Accessed: Aug. 13, 2025. [Online]. Available: https://arxiv.org/pdf/2001.08361

work page internal anchor Pith review Pith/arXiv arXiv 2020
[11]

Training Compute-Optimal Large Language Models

J. Hoffmann et al., “Training Compute-Optimal Large Language Models,” Adv Neural Inf Process Syst , vol. 35, Mar. 2022, Accessed: Aug. 13, 2025. [Online]. Available: https://arxiv.org/pdf/2203.15556

work page internal anchor Pith review Pith/arXiv arXiv 2022
[12]

An empirical analysis of compute -optimal large language model training - Google DeepMind

“An empirical analysis of compute -optimal large language model training - Google DeepMind.” Accessed: Aug. 13, 2025. [Online]. Available: https://deepmind.google/discover/blog/an-empirical-analysis-of-compute-optimal-large- language-model-training/?utm_source=chatgpt.com

work page 2025
[13]

Switch Transformers: Scaling to Trillion Parameter Models with Simple and Efficient Sparsity

W. Fedus, B. Zoph, and N. Shazeer, “Switch Transformers: Scaling to Trillion Parameter Models with Simple and Efficient Sparsity,” Journal of Machine Learning Research, vol. 23, pp. 1 –40, Jan. 2021, Accessed: Aug. 13, 2025. [Online]. Available: https://arxiv.org/pdf/2101.03961

work page internal anchor Pith review Pith/arXiv arXiv 2021
[14]

GPT-4o System Card,

OpenAI, “GPT-4o System Card,” 2024

work page 2024
[15]

Meta releases new AI model Llama 4 | Reuters

“Meta releases new AI model Llama 4 | Reuters.” Accessed: Sep. 24, 2025. [Online]. Available: https://www.reuters.com/technology/meta -releases-new-ai-model-llama-4- 2025-04-05/

work page 2025
[16]

OWASP Top 10 for Large Language Model Applications | OWASP Foundation

“OWASP Top 10 for Large Language Model Applications | OWASP Foundation.” Accessed: Aug. 13, 2025. [Online]. Available: https://owasp.org/www-project-top-10-for- large-language-model-applications/?utm_source=chatgpt.com

work page 2025
[17]

Datasheets for Datasets,

T. Gebru et al., “Datasheets for Datasets,” Commun ACM, vol. 64, no. 12, pp. 86–92, Mar. 2018, doi: 10.1145/3458723

work page doi:10.1145/3458723 2018
[18]

Data Statements for Natural Language Processing: Toward Mitigating System Bias and Enabling Better Science,

E. M. Bender and B. Friedman, “Data Statements for Natural Language Processing: Toward Mitigating System Bias and Enabling Better Science,” Trans Assoc Comput Linguist, vol. 6, pp. 587–604, Jan. 2018, doi: 10.1162/TACL_A_00041. 32

work page doi:10.1162/tacl_a_00041 2018
[19]

Ethical and social risks of harm from Language Models

L. Weidinger et al., “Ethical and social risks of harm from Language Models,” Dec. 2021, Accessed: Aug. 13, 2025. [Online]. Available: https://arxiv.org/pdf/2112.04359

work page internal anchor Pith review Pith/arXiv arXiv 2021
[20]

Training language models to follow instructions with human feedback

L. Ouyang et al., “Training language models to follow instructions with human feedback,” Adv Neural Inf Process Syst , vol. 35, Mar. 2022, Accessed: Aug. 13, 2025. [Online]. Available: https://arxiv.org/pdf/2203.02155

work page internal anchor Pith review Pith/arXiv arXiv 2022
[21]

Holistic Evaluation of Language Models (HELM)

“Holistic Evaluation of Language Models (HELM).” Accessed: Aug. 13, 2025. [Online]. Available: https://crfm.stanford.edu/helm/?utm_source=chatgpt.com

work page 2025
[22]

NIST Trustworthy and Responsible AI NIST AI 600-1 Artificial Intelligence Risk Management Framework: Generative Artificial Intelligence Profile,

G. M. Raimondo et al., “NIST Trustworthy and Responsible AI NIST AI 600-1 Artificial Intelligence Risk Management Framework: Generative Artificial Intelligence Profile,” 2024, doi: 10.6028/NIST.AI.600-1

work page doi:10.6028/nist.ai.600-1 2024
[23]

Taxonomy of Risks posed by Language Models,

L. Weidinger et al., “Taxonomy of Risks posed by Language Models,” ACM International Conference Proceeding Series , vol. 22, pp. 214 –229, Jun. 2022, doi: 10.1145/3531146.3533088;CSUBTYPE:STRING:CONFERENCE

work page doi:10.1145/3531146.3533088;csubtype:string:conference 2022
[24]

Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks

P. Lewis et al., “Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks,” Adv Neural Inf Process Syst , vol. 2020-December, May 2020, Accessed: Aug. 13, 2025. [Online]. Available: https://arxiv.org/pdf/2005.11401

work page internal anchor Pith review Pith/arXiv arXiv 2020
[25]

Making AI Less ‘Thirsty’: Uncovering and Addressing the Secret Water Footprint of AI Models,

P. Li, J. Yang, M. A. Islam, and S. Ren, “Making AI Less ‘Thirsty’: Uncovering and Addressing the Secret Water Footprint of AI Models,” Mar. 2025, Accessed: Aug. 13,

work page 2025
[27]

Article 55: Obligations for Providers of General-Purpose AI Models with Systemic Risk | EU Artificial Intelligence Act

“Article 55: Obligations for Providers of General-Purpose AI Models with Systemic Risk | EU Artificial Intelligence Act.” Accessed: Aug. 13, 2025. [Online]. Available: https://artificialintelligenceact.eu/article/55/?utm_source=chatgpt.com

work page 2025
[28]

Frontier AI Safety Commitments, AI Seoul Summit 2024 - GOV .UK

“Frontier AI Safety Commitments, AI Seoul Summit 2024 - GOV .UK.” Accessed: Aug. 13, 2025. [Online]. Available: https://www.gov.uk/government/publications/frontier -ai- safety-commitments-ai-seoul-summit-2024?utm_source=chatgpt.com

work page 2024
[29]

Universal and Transferable Adversarial Attacks on Aligned Language Models

A. Zou, Z. Wang, N. Carlini, M. Nasr, J. Z. Kolter, and M. Fredrikson, “Universal and Transferable Adversarial Attacks on Aligned Language Models,” Jul. 2023, Accessed: Aug. 13, 2025. [Online]. Available: https://arxiv.org/pdf/2307.15043

work page internal anchor Pith review Pith/arXiv arXiv 2023
[30]

The PRISMA 2020 statement: An updated guideline for reporting systematic reviews,

M. J. Page et al. , “The PRISMA 2020 statement: An updated guideline for reporting systematic reviews,” BMJ, vol. 372, Mar. 2021, doi: 10.1136/BMJ.N71

work page doi:10.1136/bmj.n71 2020
[31]

On Protecting the Data Privacy of Large Language Models (LLMs): A Survey,

B. Yan et al., “On Protecting the Data Privacy of Large Language Models (LLMs): A Survey,” Mar. 2024, Accessed: Jul. 09, 2025. [Online]. Available: https://arxiv.org/pdf/2403.05156

work page arXiv 2024
[32]

Leaner Training, Lower Leakage: Revisiting Memorization in LLM Fine-Tuning with LoRA,

F. Wang and B. Li, “Leaner Training, Lower Leakage: Revisiting Memorization in LLM Fine-Tuning with LoRA,” Jun. 2025, Accessed: Aug. 04, 2025. [Online]. Available: https://arxiv.org/pdf/2506.20856

work page arXiv 2025
[33]

Unmasking and quantifying racial bias of large language models in medical report generation,

Y . Yang, X. Liu, Q. Jin, F. Huang, and Z. Lu, “Unmasking and quantifying racial bias of large language models in medical report generation,” Communications Medicine, vol. 4, no. 1, pp. 1 –6, Dec. 2024, doi: 10.1038/S43856 -024-00601- Z;SUBJMETA=308,692,700;KWRD=HEALTH+CARE,MEDICAL+RESEARCH. 33

work page doi:10.1038/s43856 2024
[34]

Participation in the age of foundation models,

H. Suresh, E. Tseng, M. Young, M. L. Gray, E. Pierson, and K. Levy, “Participation in the age of foundation models,” 2024 ACM Conference on Fairness, Accountability, and Transparency, F AccT 2024 , vol. 1, pp. 1609 –1621, May 2024, doi: 10.1145/3630106.3658992

work page doi:10.1145/3630106.3658992 2024
[35]

Right to be Forgotten in the Era of Large Language Models: Implications, Challenges, and Solutions,

D. Zhang et al., “Right to be Forgotten in the Era of Large Language Models: Implications, Challenges, and Solutions,” AI and Ethics, vol. 5, no. 3, pp. 2445 –2454, Jul. 2023, doi: 10.1007/s43681-024-00573-9

work page doi:10.1007/s43681-024-00573-9 2023
[36]

Redefining boundaries in innovation and knowledge domains: Investigating the impact of generative artificial intelligence on copyright and intellectual property rights,

A. S. Al-Busaidi et al., “Redefining boundaries in innovation and knowledge domains: Investigating the impact of generative artificial intelligence on copyright and intellectual property rights,” Journal of Innovation & Knowledge, vol. 9, no. 4, p. 100630, Oct. 2024, doi: 10.1016/J.JIK.2024.100630

work page doi:10.1016/j.jik.2024.100630 2024
[37]

Earth Science Data Repositories: Implementing the CARE Principles,

M. O’brien et al., “Earth Science Data Repositories: Implementing the CARE Principles,” Data Sci J, vol. 23, 2024, doi: 10.5334/DSJ-2024-037

work page doi:10.5334/dsj-2024-037 2024
[38]

Privacy Auditing in Differential Private Machine Learning: The Current Trends,

I. Namatevs, K. Sudars, A. Nikulins, and K. Ozols, “Privacy Auditing in Differential Private Machine Learning: The Current Trends,” Applied Sciences 2025, Vol. 15, Page 647, vol. 15, no. 2, p. 647, Jan. 2025, doi: 10.3390/APP15020647

work page doi:10.3390/app15020647 2025
[39]

Risk Taxonomy, Mitigation, and Assessment Benchmarks of Large Language Model Systems,

T. Cui et al. , “Risk Taxonomy, Mitigation, and Assessment Benchmarks of Large Language Model Systems,” Jan. 2024, Accessed: Aug. 04, 2025. [Online]. Available: https://arxiv.org/pdf/2401.05778

work page arXiv 2024
[40]

Learning Fair Representations through Uniformly Distributed Sensitive Attributes,

P. J. Kenfack, A. R. Rivera, A. M. Khan, and M. Mazzara, “Learning Fair Representations through Uniformly Distributed Sensitive Attributes,” Proceedings - 2023 IEEE Conference on Secure and Trustworthy Machine Learning, SaTML 2023, pp. 58–67, 2023, doi: 10.1109/SATML54575.2023.00014

work page doi:10.1109/satml54575.2023.00014 2023
[41]

A Survey on Concept Factorization: From Shallow to Deep Representation Learning,

Z. Zhang, Y . Zhang, M. Xu, L. Zhang, Y . Yang, and S. Yan, “A Survey on Concept Factorization: From Shallow to Deep Representation Learning,” Inf Process Manag, vol. 58, no. 3, Jul. 2020, doi: 10.1016/j.ipm.2021.102534

work page doi:10.1016/j.ipm.2021.102534 2020
[42]

Benchmark Data Contamination of Large Language Models: A Survey,

C. Xu, S. Guan, D. Greene, and M.-T. Kechadi, “Benchmark Data Contamination of Large Language Models: A Survey,” Jun. 2024, Accessed: Aug. 04, 2025. [Online]. Available: https://arxiv.org/pdf/2406.04244

work page arXiv 2024
[43]

The Impossibility of Fair LLMs,

J. Anthis, K. Lum, M. Ekstrand, A. Feller, A. D’Amour, and C. Tan, “The Impossibility of Fair LLMs,” May 2024, Accessed: Aug. 04, 2025. [Online]. Available: https://arxiv.org/pdf/2406.03198

work page arXiv 2024
[44]

Copyright in AI Training Data: A Human-Centered Approach,

D. W. Opderbeck, “Copyright in AI Training Data: A Human-Centered Approach,” SSRN Electronic Journal, Dec. 2023, doi: 10.2139/SSRN.4679299

work page doi:10.2139/ssrn.4679299 2023
[45]

Detecting Personally Identifiable Information Through Natural Language Processing: A Step Forward,

L. Mainetti and A. Elia, “Detecting Personally Identifiable Information Through Natural Language Processing: A Step Forward,” Applied System Innovation 2025, Vol. 8, Page 55, vol. 8, no. 2, p. 55, Apr. 2025, doi: 10.3390/ASI8020055

work page doi:10.3390/asi8020055 2025
[46]

On the Diversity of Synthetic Data and its Impact on Training Large Language Models,

H. Chen et al., “On the Diversity of Synthetic Data and its Impact on Training Large Language Models,” Oct. 2024, Accessed: Aug. 04, 2025. [Online]. Available: https://arxiv.org/pdf/2410.15226 34

work page arXiv 2024
[47]

Unveiling security, privacy, and ethical concerns of ChatGPT,

X. Wu, R. Duan, and J. Ni, “Unveiling security, privacy, and ethical concerns of ChatGPT,” Journal of Information and Intelligence, vol. 2, no. 2, pp. 102–115, Mar. 2024, doi: 10.1016/j.jiixd.2023.10.007

work page doi:10.1016/j.jiixd.2023.10.007 2024
[48]

A Framework for Understanding Sources of Harm throughout the Machine Learning Life Cycle,

H. Suresh and J. V . Guttag, “A Framework for Understanding Sources of Harm throughout the Machine Learning Life Cycle,” ACM International Conference Proceeding Series , Dec. 2021, doi: 10.1145/3465416.3483305

work page doi:10.1145/3465416.3483305 2021
[49]

Reconciling the contrasting narratives on the environmental impact of large language models,

S. Ren, B. Tomlinson, R. W. Black, and A. W. Torrance, “Reconciling the contrasting narratives on the environmental impact of large language models,” Sci Rep, vol. 14, no. 1, pp. 1 –8, Dec. 2024, doi: 10.1038/S41598 -024-76682- 6;SUBJMETA=4066,4069,685,704,844;KWRD=ENERGY+EFFICIENCY ,SUSTAINA BILITY

work page doi:10.1038/s41598 2024
[50]

Reconciling the contrasting narratives on the environmental impact of large language models,

S. Ren, B. Tomlinson, R. W. Black, and A. W. Torrance, “Reconciling the contrasting narratives on the environmental impact of large language models,” Sci Rep, vol. 14, no. 1, pp. 1–8, Dec. 2024, doi: 10.1038/S41598-024-76682-6;SUBJMETA

work page doi:10.1038/s41598-024-76682-6;subjmeta 2024
[51]

Reaching carbon neutrality requires energy -efficient training of AI,

L. Guan, “Reaching carbon neutrality requires energy -efficient training of AI,” Nature, vol. 626, no. 7997, p. 33, Feb. 2024, doi: 10.1038/D41586-024-00200-X;SUBJMETA

work page doi:10.1038/d41586-024-00200-x;subjmeta 2024
[52]

The carbon emissions of writing and illustrating are lower for AI than for humans,

B. Tomlinson, R. W. Black, D. J. Patterson, and A. W. Torrance, “The carbon emissions of writing and illustrating are lower for AI than for humans,” Sci Rep, vol. 14, no. 1, pp. 1–8, Dec. 2024, doi: 10.1038/S41598-024-54271-X;SUBJMETA

work page doi:10.1038/s41598-024-54271-x;subjmeta 2024
[53]

How Hungry is AI? Benchmarking Energy, Water, and Carbon Footprint of LLM Inference - Google Search

“How Hungry is AI? Benchmarking Energy, Water, and Carbon Footprint of LLM Inference - Google Search.” Accessed: Aug. 04, 2025. [Online]. Available: https://www.google.com/search?q=How+Hungry+is+AI%3F+Benchmarking+Energy% 2C+Water%2C+and+Carbon+Footprint+of+LLM+Inference&rlz=1C1UEAD_enUS102 4US1024&oq=How+Hungry+is+AI%3F+Benchmarking+Energy%2C+Water%2C+an...

work page 2025
[54]

Enhancing Supply Chain Visibility with Generative AI: An Exploratory Case Study on Relationship Prediction in Knowledge Graphs,

G. Zheng and A. Brintrup, “Enhancing Supply Chain Visibility with Generative AI: An Exploratory Case Study on Relationship Prediction in Knowledge Graphs,” Dec. 2024, Accessed: Aug. 04, 2025. [Online]. Available: https://arxiv.org/pdf/2412.03390

work page arXiv 2024
[55]

Making AI Less ‘Thirsty’: Uncovering and Addressing the Secret Water Footprint of AI Models,

P. Li, J. Yang, M. A. Islam, and S. Ren, “Making AI Less ‘Thirsty’: Uncovering and Addressing the Secret Water Footprint of AI Models,” Apr. 2023, Accessed: Aug. 04,

work page 2023
[56]

Available: https://arxiv.org/pdf/2304.03271

[Online]. Available: https://arxiv.org/pdf/2304.03271

work page arXiv
[57]

Quantifying the Energy Consumption and Carbon Emissions of LLM Inference via Simulations,

M. Özcan, P. Wiesner, P. Weiß, and O. Kao, “Quantifying the Energy Consumption and Carbon Emissions of LLM Inference via Simulations,” Jul. 2025, Accessed: Aug. 04,

work page 2025
[58]

Available: https://arxiv.org/pdf/2507.11417

[Online]. Available: https://arxiv.org/pdf/2507.11417

work page arXiv
[59]

Life-Cycle Emissions of AI Hardware: A Cradle-To-Grave Approach and Generational Trends,

I. Schneider et al., “Life-Cycle Emissions of AI Hardware: A Cradle-To-Grave Approach and Generational Trends,” Feb. 2025, Accessed: Aug. 04, 2025. [Online]. Available: https://arxiv.org/pdf/2502.01671

work page arXiv 2025
[60]

Carbon- and Precedence-Aware Scheduling for Data Processing Clusters,

A. Lechowicz, R. Shenoy, N. Bashir, M. Hajiesmaili, A. Wierman, and C. Delimitrou, “Carbon- and Precedence-Aware Scheduling for Data Processing Clusters,” Feb. 2025, Accessed: Aug. 04, 2025. [Online]. Available: https://arxiv.org/pdf/2502.09717 35

work page arXiv 2025
[61]

(PDF) Ghost Work in Modern Business: Opportunities and Challenges

“(PDF) Ghost Work in Modern Business: Opportunities and Challenges.” Accessed: Aug. 04, 2025. [Online]. Available: https://www.researchgate.net/publication/390517385_Ghost_Work_in_Modern_Busines s_Opportunities_and_Challenges

work page arXiv 2025
[62]

The Glamorisation of Unpaid Labour: AI and its Influencers,

N. M. Nwachukwu, J. S. Roberts, and L. N. Montoya, “The Glamorisation of Unpaid Labour: AI and its Influencers,” Jul. 2023, Accessed: Aug. 04, 2025. [Online]. Available: https://arxiv.org/pdf/2308.02399

work page arXiv 2023
[63]

Playing Devil’s Advocate: Unmasking Toxicity and Vulnerabilities in Large Vision -Language Models,

A. Erol, T. Padhi, A. Saha, U. Kursuncu, and M. E. Aktas, “Playing Devil’s Advocate: Unmasking Toxicity and Vulnerabilities in Large Vision -Language Models,” Jan. 2025, Accessed: Aug. 04, 2025. [Online]. Available: https://arxiv.org/pdf/2501.09039

work page arXiv 2025
[64]

NBER WORKING PAPER SERIES GENERATIVE AI AND FIRM V ALUES,

A. L. Eisfeldt, G. Schubert, and M. Ben Zhang, “NBER WORKING PAPER SERIES GENERATIVE AI AND FIRM V ALUES,” 2023, Accessed: Aug. 04, 2025. [Online]. Available: http://www.nber.org/papers/w31222

work page 2023
[65]

On pipelines, readiness and annotative labour: Political geographies of AI and data infrastructures in Africa,

K. Holden and M. Harsh, “On pipelines, readiness and annotative labour: Political geographies of AI and data infrastructures in Africa,” Polit Geogr, vol. 113, p. 103150, Aug. 2024, doi: 10.1016/J.POLGEO.2024.103150

work page doi:10.1016/j.polgeo.2024.103150 2024
[66]

Bias and Fairness in Large Language Models: A Survey,

I. O. Gallegos et al. , “Bias and Fairness in Large Language Models: A Survey,” Computational Linguistics , vol. 50, no. 3, pp. 1097 –1179, Sep. 2023, doi: 10.1162/coli_a_00524

work page doi:10.1162/coli_a_00524 2023
[67]

Sociodemographic biases in medical decision making by large language models,

M. Omar et al., “Sociodemographic biases in medical decision making by large language models,” Nat Med, vol. 31, no. 6, pp. 1873 –1881, Jun. 2025, doi: 10.1038/S41591 -025- 03626- 6;SUBJMETA=3935,478,692,700;KWRD=MEDICAL+ETHICS,PUBLIC+HEALTH

work page doi:10.1038/s41591 2025
[68]

Could a Conversational AI Identify Offensive Language?,

D. A. da Silva et al. , “Could a Conversational AI Identify Offensive Language?,” Information 2021, Vol. 12, Page 418 , vol. 12, no. 10, p. 418, Oct. 2021, doi: 10.3390/INFO12100418

work page doi:10.3390/info12100418 2021
[69]

‘They only care to show us the wheelchair’: Disability Representation in Text -to-Image AI Models,

K. A. Mack, R. Qadri, R. Denton, S. K. Kane, and C. L. Bennett, “‘They only care to show us the wheelchair’: Disability Representation in Text -to-Image AI Models,” Conference on Human Factors in Computing Systems - Proceedings, vol. 23, no. 24, May 2024, doi: 10.1145/3613904.3642166/SUPPL_FILE/3613904.3642166-TALK-VIDEO.VTT

work page doi:10.1145/3613904.3642166/suppl_file/3613904.3642166-talk-video.vtt 2024
[70]

Surprising gender biases in GPT,

R. A. Fulgu and V . Capraro, “Surprising gender biases in GPT,” Computers in Human Behavior Reports, vol. 16, p. 100533, Dec. 2024, doi: 10.1016/J.CHBR.2024.100533

work page doi:10.1016/j.chbr.2024.100533 2024
[71]

Risks of Cultural Erasure in Large Language Models,

R. Qadri, A. M. Davani, K. Robinson, and V . Prabhakaran, “Risks of Cultural Erasure in Large Language Models,” Jan. 2025, Accessed: Aug. 04, 2025. [Online]. Available: https://arxiv.org/pdf/2501.01056

work page arXiv 2025
[72]

Mitigate One, Skew Another? Tackling Intersectional Biases in Text-to- Image Models,

P. Shukla et al., “Mitigate One, Skew Another? Tackling Intersectional Biases in Text-to- Image Models,” May 2025, Accessed: Aug. 04, 2025. [Online]. Available: https://arxiv.org/pdf/2505.17280

work page arXiv 2025
[73]

How Far Can We Extract Diverse Perspectives from Large Language Models?,

S. A. Hayati, M. Lee, D. Rajagopal, and D. Kang, “How Far Can We Extract Diverse Perspectives from Large Language Models?,” EMNLP 2024 - 2024 Conference on 36 Empirical Methods in Natural Language Processing, Proceedings of the Conference, pp. 5336–5366, Nov. 2023, doi: 10.18653/v1/2024.emnlp-main.306

work page doi:10.18653/v1/2024.emnlp-main.306 2024
[74]

Sometimes the Model doth Preach: Quantifying Religious Bias in Open LLMs through Demographic Analysis in Asian Nations,

H. Shankar, V . S. P, T. Cavale, P. Kumaraguru, and A. Chakraborthy, “Sometimes the Model doth Preach: Quantifying Religious Bias in Open LLMs through Demographic Analysis in Asian Nations,” Mar. 2025, Accessed: Aug. 04, 2025. [Online]. Available: https://arxiv.org/pdf/2503.07510

work page arXiv 2025
[75]

Examining Identity Drift in Conversations of LLM Agents,

J. Choi, Y . Hong, M. Kim, and B. Kim, “Examining Identity Drift in Conversations of LLM Agents,” Dec. 2024, Accessed: Aug. 04, 2025. [Online]. Available: https://arxiv.org/pdf/2412.00804

work page arXiv 2024
[76]

ChatGPT Exhibits Bias Toward Developed Countries Over Developing Ones, as Indicated by a Sentiment Analysis Approach,

G. P. Georgiou, “ChatGPT Exhibits Bias Toward Developed Countries Over Developing Ones, as Indicated by a Sentiment Analysis Approach,” J Lang Soc Psychol, vol. 44, no. 1, pp. 132–141, Jan. 2025, doi: 10.1177/0261927X241298337

work page doi:10.1177/0261927x241298337 2025
[77]

Unequal V oices: How LLMs Construct Constrained Queer Narratives,

A. Ghosal, A. Gupta, and V . Srikumar, “Unequal V oices: How LLMs Construct Constrained Queer Narratives,” Jul. 2025, Accessed: Aug. 04, 2025. [Online]. Available: https://www.arxiv.org/pdf/2507.15585

work page arXiv 2025
[78]

Seeing Like an AI: How LLMs Apply (and Misapply) Wikipedia Neutrality Norms,

J. Ashkinaze, R. Guan, L. Kurek, E. Adar, C. Budak, and E. Gilbert, “Seeing Like an AI: How LLMs Apply (and Misapply) Wikipedia Neutrality Norms,” Proceedings of Preprint, vol. 1, Jul. 2024, doi: XXXXXXX.XXXXXXX

work page 2024
[79]

Quantifying Multilingual Performance of Large Language Models Across Languages,

Z. Li et al., “Quantifying Multilingual Performance of Large Language Models Across Languages,” Proceedings of the AAAI Conference on Artificial Intelligence , vol. 39, no. 27, pp. 28186–28194, Apr. 2024, doi: 10.1609/aaai.v39i27.35038

work page doi:10.1609/aaai.v39i27.35038 2024
[80]

Diminishing Stereotype Bias in Image Generation Model using Reinforcemenlent Learning Feedback,

X. Chen and V . Foussereau, “Diminishing Stereotype Bias in Image Generation Model using Reinforcemenlent Learning Feedback,” Jun. 2024, Accessed: Aug. 04, 2025. [Online]. Available: https://arxiv.org/pdf/2407.09551

work page arXiv 2024
[81]

Targeted Data Augmentation for bias mitigation,

A. Mikołajczyk-Bareła, M. Ferlin, and M. Grochowski, “Targeted Data Augmentation for bias mitigation,” Aug. 2023, Accessed: Aug. 04, 2025. [Online]. Available: https://arxiv.org/pdf/2308.11386

work page arXiv 2023
[82]

Does RAG Introduce Unfairness in LLMs? Evaluating Fairness in Retrieval -Augmented Generation Systems,

X. Wu, S. Li, H. T. Wu, Z. Tao, and Y . Fang, “Does RAG Introduce Unfairness in LLMs? Evaluating Fairness in Retrieval -Augmented Generation Systems,” Proceedings - International Conference on Computational Linguistics, COLING , vol. Part F206484 -1, pp. 10021 –10036, Sep. 2024, Accessed: Aug. 04, 2025. [Online]. Available: https://arxiv.org/pdf/2409.19804

work page arXiv 2024
[83]

Assessing the potential of GPT-4 to perpetuate racial and gender biases in health care: a model evaluation study,

T. Zack et al., “Assessing the potential of GPT-4 to perpetuate racial and gender biases in health care: a model evaluation study,” Lancet Digit Health , vol. 6, no. 1, pp. e12 –e22, Jan. 2024, doi: 10.1016/S2589-7500(23)00225-X

work page doi:10.1016/s2589-7500(23)00225-x 2024

Showing first 80 references.