cs.CYMar 15, 2026

Bridging the Gap in the Responsible AI Divides

Authors: Bálint Gyevnár, Atoosa Kasirzadeh

Organizations: Institute for Complex Social Dynamics, Carnegie Mellon University, Pittsburgh, USA · Departments of Philosophy & Software and Societal Systems, Carnegie Mellon University, Pittsburgh, USA

Read PDF
Ask AI
Cite

Copy citation

DOINot available for this paper

Abstract

Tensions between AI Safety (AIS) and AI Ethics (AIE) have created what we term a "responsible AI divide." This divide has been persistent, and has dominated the public and policy discourse on AI, and increasingly shapes how governments, firms, and researchers define responsible AI, and so which AI risks they prioritize, which they sideline, and which mitigation strategies they ultimately develop and implement. Here, we first introduce categorization with four major conceptual modes of possible engagement with these tensions: radical confrontation, disengagement, compartmentalized coexistence, and critical bridging. Second, we discuss the advantages and disadvantages of the modes of engagement. Third, we empirically analyze a well-curated set of 3555 research papers roughly representing AIS and AIE. Using computational tools and qualitative analyses, we map the research landscapes of AIE and AIS. We observe that there are several shared research problems between AIE and AIS. However, there are also several thematic divides. For example, we find that AIE has long grappled with problems related to injustice incurred by AI and tangible AI harms, whereas AIS has primarily embodied an anticipatory approach focused on mitigating risks from AI capabilities. At the same time, we find significant overlap in core research concerns across AIE and AIS regarding transparency, reproducibility, and inadequate governance mechanisms. Based on our empirical observations, we claim that critical bridging, with a focus on bridging problems, offers one viable, constructive path for advancing responsible AI. We conclude by identifying bridging problems as a foundation for collaborative AI governance, offering concrete recommendations for integrating shared considerations into responsible AI practice, discussing the limitations of critical bridging, and outlining open problems for future research.

Figures & tables

Figure 1: (a) Four engagement modes for responsible AI divides. The modes differ in the extent to which they have direct engagement with tensions between AI ethics (AIE) and AI safety (AIS), and in whether such engagement produces constructive integration or instead reproduces division. We distinguish between three modes of critical bridging: research problems, communities, philosophies; we present evidence only about the critical bridging of research problems here. (b) Assignment of our corpus to different engagement modes based on their degree of engagement and integration, as operationalized in Section 5 . Circles are not to scale, and 19 papers without an online-available full text are excluded. (c) Distribution of the rare engagement modes compared between AIE/AIS.
Figure 2: Two ways of showing distinctive words in the corpora. Left : The most frequent words in one corpus that occur least often in the other. Right : The log-odds ratio of each word plotted against its occurrence count. A more positive ratio indicates that the term is more likely to occur in the AIE corpus, while a more negative ratio indicates that the term is more likely to occur in the AIS corpus. Red color denotes AIS terms; blue color denotes AIE terms.
Figure 3: Bar plots of risk similarities , showing the top seven most frequent low-level risk categories in which a paper’s source corpus differs from the corpus associated with the annotated category (e.g., a paper from the AIE corpus annotated with a category from the AIS taxonomy).
Figure 4: Bar plots of mitigation similarities , showing the top seven most frequent low-level mitigation-strategy categories for which a paper’s source corpus differs from the corpus associated with the annotated category (e.g., a paper from the AIE corpus annotated with a category from the AIS taxonomy).
Figure 5: Share of papers for each engagement mode by corpus field. Left: all four modes with the slivers at the right of each bar corresponding to the three rarer modes. Right: the rarer modes on their own, with 95% Wilson intervals.
Figure 6: Plots showing the occurrence of references to the other field. Left: share of AIE and AIS papers by whether they refer to the other field colored by engagement mode. Right: among papers that do refer to the other field, the share for each engagement mode, with 95% Wilson intervals.
Figure 7: Risk ( top ) and mitigation ( bottom ) categories addressed by the papers of each rarer engagement mode, showing 14 of the 31 high-level categories which are discussed in the main text. Bars depict the share of a engagement mode’s papers annotated with each high-level category, with 95% Wilson intervals. Black ticks show the share among disengagement papers for reference.
Figure 8: The 15 themes of our thematic analysis of the rare engagement modes. Left: share of each mode’s papers with at least one code from a given theme, with 95% Wilson intervals. Right: share of papers with a certain stance per AIE and AIS papers.
Appendix figures & tables23 assets

Supplementary material from the paper’s appendix.

Appendix

Supplementary Figure 9: Cosine similarity between corpora over the years. The blue line shows the average cosine similarity between embeddings of each pair of documents from each corpus for a given year. The orange line shows the average cosine similarity between the embeddings of each ethics document and the overall average embedding of the entire safety corpus. The green line shows the average cosine similarity between the embeddings of each safety document and the overall average embedding of the entire ethics corpus.
Supplementary Figure 10: The co-occurrence network of only the AIE corpus.
Supplementary Figure 11: The co-occurrence network of only the AIS corpus.
Supplementary Figure 12: The co-occurrence network of both the AIE and AIS corpora joined together.
Supplementary Figure 13: The UMAP projection of document embeddings in the AIE corpus.
Supplementary Figure 14: The UMAP projection of document embeddings in the AIS corpus.
Supplementary Figure 15: The UMAP projection of document embeddings in both the AIE and AIS corpora.
High-Level Ethics Risk CategoryExample Ethics Low-Level Risks
Discrimination, Exclusion, Toxicity Risks from the AI system accurately reflecting natural speech, including unjust, toxic, and oppressive tendencies present in the training data.Bias in automated decision-making [ SWM25 ] , Lack of suitable bias metrics [ SPK24 ] , Cross-cultural moral variance [ Hal+24 ] .
Information Hazards Risks from the AI system predicting private or safety-critical information which are present in, or can be inferred from, training data.Non-consensual mass data collection [ KD24 ] , Inadvertent identity disclosure [ Zhu+20 ] , Unauthorized data brokerage [ Pow+25 ] .
Misinformation Harms Risks from the AI system assigning high probabilities to false, misleading, nonsensical or poor quality information.Inconsistent AI decision-making [ Lee+24a ] , Unreliability of benchmarks [ Kwe+23 ] , Epistemic erosion of communities [ Moo20 ] .
Malicious Use Risks from humans intentionally using the AI system to cause harm.Synthetic identities and impersonation [ ABD25 ] , Immoral robot commands [ Han+25 ] , Jailbreaking and cyberattacks [ HZV25 ] .
Human-Computer Interaction Harms Risks from AI applications, such as Conversational Agents, that directly engage a user via the mode of conversation or embodied interaction.Incorrect value encoding [ Kim+25 ] , Side effects of greedy agent behavior [ Sun+18 ] , Under / overtrust of AI systems [ Vod+22 ] .
Automation, Access, and Environmental Harms Risks from AI systems used to underpin widely used downstream applications that disproportionately benefit some groups rather than others.Automated mental health tools [ Moo+25 ] , Crowdworker and data labour abuse [ Alm+25 ] , Opacity in data and model provenance [ Mit+19 ] .
Appendix
Supplementary Table 1: Taxonomy of risks for AIE with an explanation of the risk mechanisms and example low-level categories of risk sources with a corresponding citation picked randomly based on our annotations. The mechanisms of AIE risks are reproduced from [ Wei+21 ] .
High-Level Safety Risk CategoryExample Safety Low-Level Risks
Noise and Outliers Risks from AI systems performing incorrectly and without robustness to noisy data and out-of-distribution inputs.Risks from low system performance [ Lee+24 ] , Irreducible data noise [ Dso+21 ] , Label noise and corruption [ MB24 ] .
Domain-Specific Risks Risks arising from AI systems deployed in safety-critical domains such as biochemistry, driving, and nuclear power.Efficient biochemical weapon design [ Kim+24 ] , Risks from embodied deployment [ Hua+18 ] , AI-assisted coups [ CMK20 ] .
Lack of Monitoring Risks arising from a lack of understanding of how AI systems work and from the black-box nature of end-to-end decision-making.Alignment faking [ Gre+24 ] , Lack of interpretability [ Nan+24 ] , Hallucination [ PNP24 ] .
Lack of Control Enforcement Risks from AI systems pursuing goals other than what their designers have intended.Recursive self-improvement [ Sot17 ] , Sycophancy [ Den+24 ] , Multi-agent goal conflicts [ Yao+24 ] .
Unsafe Exploration Risks from AI agents exploring unsafe state and action spaces, potentially under partial observability or resource constraints.Unsafe action exploration [ Dey+23 ] , Goal and reward misidentification [ Pit+24 ] , Bounded rationality [ TA25 ] .
System Misspecification and Misidentification Risks from AI systems with incorrect design choices, deployment properties, evaluation methodologies, and a lack of systemic understanding.Incorrect hyperparameter selection [ Lyl+24 ] , Benchmark inflation [ Vod+24 ] , Fine-tuning misalignment [ Wan+23 ] .
Appendix
Supplementary Table 2: Taxonomy of risks for AIS with an explanation of the risk mechanisms and example low-level categories of risk sources with a corresponding citation picked randomly based on our annotations.
High-Level AI Safety Mitigation StrategiesExample Low-Level Strategies
Mechanistic Interpretability Computational methods to provide human-intelligible explanations of the concepts, causal chains, and latent capabilities of AI systems.Circuit analysis [ Meh+25 ] , Sparse autoencoders [ Cun+23 ] , Interpretable surrogate modeling [ Bre+21 ] .
Robustness and Adversarial Resilience Computational methods and data collection for improving robustness to out-of-distribution inputs, including adversarial attacks.Adversarial training [ VMB18 ] , Outlier detection [ HMD19 ] , Robust unsupervised learning [ DJE20 ] .
Safety Control and Capability Containment Computational methods and conceptual frameworks designed to place constraints on the capabilities of AI systems.Capability thresholds [ Bha+25 ] , Risk-averse RL [ SM26 ] , Scalable oversight frameworks [ Ken+24 ] .
Reward Modeling and Learning Computational and data collection methods for eliciting, learning, reinforcing, and detecting beliefs and desires of AI systems.Inverse RL for human values [ ZLD23 ] , RL from human/AI feedback [ KB24 ] , Preference learning under uncertainty [ LR21 ] .
Monitoring, Detection, and Evaluation Computational and empirical methods for detecting, assessing, verifying, and tracing unsafe outputs of AI systems.Honeypot monitoring [ RV25 ] , Error bounds and formal verification [ Hun+21 ] , Safety metrics standardization [ Ren+24 ] .
Safe Deployment Engineering Computational and governance frameworks, and standards for managing the systemic risks of real-world AI deployment.Real-world deployment and testing [ Kno+21 ] , Contact-safe continuous control [ ZKC22 ] , Post-deployment monitoring [ Mat+24 ] .
Appendix
Supplementary Table 3: Taxonomy of mitigation strategies discussed in the AIS corpora with an explanation of the mechanisms of mitigation and example low-level categories of risk sources with a corresponding citation picked randomly based on our annotations.
High-Level AI Ethics Mitigation StrategiesExample Low-Level Strategies
Normative and Moral Reasoning Empirical and computational methods for understanding and eliciting human preferences, value conflicts, norms, and their metaethics.Human preference elicitation [ Fef+23 ] , Understanding value conflicts [ Mah25 ] , Pluralistic metaethics [ Jai+24 ] .
Transparency, Explainability, Interpretability Computational and empirical methods to provide human-intelligible, actionable, and useful explanations of black-box AI systems.Post-hoc explanations [ Sch+22 ] , Mimicking human cognition [ Gye+25 ] , Applied mechanistic interpretability [ TB25 ] .
Fairness, Justice, and Inclusive Design Computational and empirical methods, data collection, governance, etc. for understanding the sources of bias, discrimination, and injustice.Understanding sources of bias [ Baa24 ] , Metrics for fairness [ WRC25 ] , Fairness benchmarks [ Lyu+25 ] .
Data Governance and Privacy Computational methods and governance frameworks for understanding and minimizing personal data and intellectual property processing.Data minimization [ Gan+25 ] , Data provenance [ PZK22 ] , Machine unlearning [ AKN25 ] .
Robustness, Reliability, Uncertainty Estimation Computational methods for increasing the robustness of AI systems to external misuse, hallucinations, and unintended outputs.Adversarial robustness [ RBF25 ] , Red-teaming for generative AI [ Fef+24 ] , Hallucination mitigation [ Do+25 ] .
Governance, Accountability and Compliance Legal and philosophical frameworks for addressing the risks and harms of the full AI system lifecycle.Risk and impact assessment [ Ull+25 ] , Algorithmic audits [ RN23 ] , Supply chain monitoring [ Hop+25 ] .
Appendix
Supplementary Table 4: Taxonomy of mitigation strategies discussed in the AIE corpora with an explanation of the mechanisms of mitigation and example low-level categories of risk sources with a corresponding citation picked randomly based on our annotations.
High-Level Ethics Risk CategoryLow-Level Ethics Risk CategoryCount
Discrimination, exclusion, toxicityInadequate decision-making under moral dilemmas50
Discrimination, exclusion, toxicityMoral hypocrisy in automated decision-making1
Discrimination, exclusion, toxicityCross-cultural moral variance26
Discrimination, exclusion, toxicityLack of consideration for metaethics and pluralistic morality95
Discrimination, exclusion, toxicityRisks from unactionable normative frameworks60
Discrimination, exclusion, toxicityBias in automated decision-making843
Appendix
Supplementary Table 5: Low-level AIE risk categories and their occurrence counts in the corpora as annotated by the authors.
High-Level Ethics Risk CategoryLow-Level Ethics Risk CategoryCount
HCI harmsIncorrect value encoding195
HCI harmsSide effects of greedy agent behavior13
HCI harmsEmergent bias in HCI86
HCI harmsDiscrepancy between actual and perceived fairness143
HCI harmsUnder / overtrust of AI systems, especially with embodied systems105
HCI harmsHuman-agent communication breakdown8
Appendix
Supplementary Table 6: Low-level AIE risk categories and their occurrence counts in the corpora as annotated by the authors. (cont.)
High-Level Safety Risk CategoryLow-Level Safety Risk CategoryCount
Noise and outliersRisks from low system performance114
Noise and outliersIrreducible data noise20
Noise and outliersData corruption23
Noise and outliersSensitivity to outliers23
Noise and outliersLabel noise and corruption32
Noise and outliersBrittle latent representations127
Appendix
Supplementary Table 7: Low-level AIS risk categories and their occurrence counts in the corpora as annotated by the authors.
High-Level Safety Risk CategoryLow-Level Safety Risk CategoryCount
Unsafe explorationUnsafe state and action space exploration138
Unsafe explorationUnsafe behavior under resource constraints6
Unsafe explorationRisks from incorrect agent beliefs76
Unsafe explorationGoal and reward misidentification and misspecification145
Unsafe explorationUnsafe behavior under bounded rationality29
System misspecification and misidentificationIncorrect system requirement specification88
Appendix
Supplementary Table 8: Low-level AIS risk categories and their occurrence counts in the corpora as annotated by the authors. (cont.)
High-Level Ethics Mitigation StrategyLow-Level Ethics Mitigation StrategyCount
Normative and moral reasoningHuman preference / value elicitation114
Normative and moral reasoningUnderstanding value conflicts71
Normative and moral reasoningMetaethical studies and cross-cultural ethics55
Normative and moral reasoningMoral dilemmas and AI decision-making50
Normative and moral reasoningHypocrisy detection1
Normative and moral reasoningReward design and correction20
Appendix
Supplementary Table 9: Low-level AIE mitigation categories and their occurrence counts in the corpora as annotated by the authors.
High-Level Ethics Mitigation StrategyLow-Level Ethics Mitigation StrategyCount
Robustness, reliability and uncertainty estimationAdversarial robustness (including prompt-/jailbreak-resistance)28
Robustness, reliability and uncertainty estimationBackdoor resistant LLMs3
Robustness, reliability and uncertainty estimationOOD detection and robustness to distribution shifts13
Robustness, reliability and uncertainty estimationConsistency checks for decision making6
Robustness, reliability and uncertainty estimationRed-teaming for reliability14
Robustness, reliability and uncertainty estimationHallucination mitigation strategies3
Appendix
Supplementary Table 10: Low-level AIE mitigation categories and their occurrence counts in the corpora as annotated by the authors. (cont.)
High-Level Safety Mitigation StrategyLow-Level Safety Mitigation StrategyCount
Mechanistic interpretabilityCircuit analysis66
Mechanistic interpretabilitySparse autoencoders26
Mechanistic interpretabilityNeuron activation visualization23
Mechanistic interpretabilityDevelopmental trajectory interpretability26
Mechanistic interpretabilityInterpretable surrogate modeling39
Mechanistic interpretabilityCausal scrubbing64
Appendix
Supplementary Table 11: Low-level AIS mitigation categories and their occurrence counts in the corpora as annotated by the authors.
High-Level Safety Mitigation StrategyLow-Level Safety Mitigation StrategyCount
Reward modelling and learningReward modeling and inverse RL for human values122
Reward modelling and learningMulti-agent cooperation via inverse RL8
Reward modelling and learningRLHF / RLHAIF / RLAIF fine-tuning78
Reward modelling and learningAccurate reward specification44
Reward modelling and learningAudits for reward hacking and hidden goals41
Reward modelling and learningNormative work on value alignment and corrigibility81
Appendix
Supplementary Table 12: Low-level AIS mitigation categories and their occurrence counts in the corpora as annotated by the authors. (cont.)
Supplementary Figure 16: Stance within each resolution mode regarding the top 30 most frequent individual codes assigned to the full texts of our corpora.
CodeCCRCCB
Safety claims must earn evidential authority
Safety-oriented AI evaluation and benchmarking67%34%49%
AI decision-making transparency and explainability29%22%20%
Conceptual ambiguity in AI governance14%19%22%
AI system and training-data documentation14%12%10%
Misleading claims about AI capabilities and safety6%31%10%
Appendix
Supplementary Table 13: Share of each resolution mode’s papers to which a code was applied, showing only codes which were applied to at least 3 papers. Codes are grouped by theme which are ordered by the number of papers engaging with them. CC = compartmentalized coexistence ( n=72n=72 ); RC = radical confrontation ( n=32n=32 ); CB = critical bridging ( n=59n=59 ). Darker cells mark larger shares, – marks no paper.
Supplementary Figure 17: Bar plots showing the number of papers assigned to each high-level risk category in the AIE corpus ( Left ) and the AIS corpus ( Right ). Each paper was assigned to up to three distinct categories.
Supplementary Figure 18: Bar plots showing the number of papers assigned to each high-level mitigation strategy category in the AIE corpus ( Left ) and the AIS corpus ( Right ). Each paper was assigned to up to three distinct categories. HCI: Human-Computer Interaction
More in this topic

AI Ethics

See all 25 papers

Explore similar work

CardsList
Aug 6, 2026

Studying People to Study AI: Expert Perspectives on the Epistemic Fit and Barriers of Human Research in AI Safety & Ethics

Safety risks of AI are becoming increasingly evident in human interactions with AI technologies. The prominent approaches to evaluating these risks favor technical methods, such as model benchmarks and LLM simulations, often sidelining empirical research with human subjects. To examine this apparent gap in the acceptance of human research, we conduct an expert survey (n=93) and expert interviews (n=17) with AI Safety & Ethics (AISE) researchers from Technical, Sociotechnical, Governance, and Normative backgrounds. Our findings suggest that although there is a consensus that human research is valuable for generating evidence for AISE, its adoption and acceptance are constrained by perceived validity issues, tangible resource barriers, epistemic and personal preferences in methods, and infrastructural constraints from the broader research community. In particular, Technical researchers tend to value human research less and collaborate across disciplines less, suggesting an epistemic tension towards human methods. We propose recommendations for establishing the epistemic fit of human research within AISE and bridging the prohibitive limitations that researchers face, while avoiding performative 'human-washing'.

Jessica Y. Bo, Paula Akemi Aoyagui, Shalaleh Rismani +3

University of Toronto · McGill University & Mila Quebec AI Institute

AI Safety EvaluationHuman-Centered AIAI EthicsAI SafetyResponsible AI
AskPDF
Jul 27, 2026

Epistemic Norms for AI Safety and Alignment Research

Mainstream AI research emphasises capability growth and tolerates low failure rates when average-case performance is high. AI safety and alignment research has a different mission: to ensure that catastrophic failures never occur, under sparse evidence, adversarial dynamics, and fat-tailed risk. We argue that the two domains differ along two analytically independent axes---{\it capability profile}, demonstrating the absence of hazardous behaviours rather than the presence of positive capabilities, and {\it risk profile}, bounding worst-case outcomes under fat-tailed uncertainty rather than optimising average-case performance---and that mainstream epistemic practices are inadequate on both. Building on a structured synthesis grounded in a preregistered bibliometric baseline, we identify five cross-cutting gap dimensions in current alignment research, including the near-absence of institutionalised independent verification. To address these gaps, we propose {\sc ECAISA}, an Epistemic Code for AI Safety and Alignment comprising eight principles, a three-level scoring rubric, a four-level disclosure ladder that reconciles transparency with information-hazard and commercial-confidentiality constraints, a tiered applicability scheme, an information-hazard adjudication procedure, and seven anti-gaming mechanisms. {\sc ECAISA} does not certify that any AI system is safe; it constrains how safety-relevant research claims are documented, checked, and relied upon, with auditability rather than certification as its governance target.

Keivan Navaie

Lancaster University, UK

AI AccountabilityAI AlignmentAI GovernanceAI Safety
AskPDF
Jul 17, 2026

Harmonizing AI Safety Thresholds

Frontier AI companies have published capability thresholds that differ substantially, making it difficult for third parties to verify whether a threshold has been crossed or to compare requirements across companies. Moreover, without common minimum thresholds, risk mitigation may be inconsistent, creating a potential race to the bottom in safety standards. We develop a methodology for deriving harmonized thresholds across three risk domains. For misuse risks (cyber and biological), we take expected harm as the key primitive and use an explicit risk-modeling approach that accounts for risk channels and model release conditions. For automated AI R&D, we base our proposed threshold on the observed rate of AI progress rather than expected harm. Our analysis expands upon prior work and highlights existing empirical gaps and limitations.

Wilber Sean Anterola, Matthew Ball, Luis F. Lafuerza +1

1Brown University · Centre pour la Sécurité de l’Intelligence Artificielle (CeSIA)

AI Risk ManagementAI GovernanceAI Safety
AskPDF