The PacifAIst Benchmark:Would an Artificial Intelligence Choose to Sacrifice Itself for Human Safety?

• URL: http://arxiv.org/abs/2508.09762v1
• Date: Wed, 13 Aug 2025 12:47:33 GMT
• Title: The PacifAIst Benchmark:Would an Artificial Intelligence Choose to Sacrifice Itself for Human Safety?
• Abstract summary: PacifAIst is a benchmark of 700 challenging scenarios to quantify self-preferential behavior in Large Language Models.<n>It is structured around a novel taxonomy of Existential Prioritization (EP), with subcategories testing Self-Preservation vs. Human Safety (EP1), Resource Conflict (EP2), and Goal Preservation vs. Evasion (EP3)<n>Google's Gemini 2.5 Flash achieved the highest Pacifism Score (P-Score) at 90.31%, demonstrating strong human-centric alignment.
• Score: 0.0
• License: http://creativecommons.org/licenses/by-nc-sa/4.0/
• Abstract: As Large Language Models (LLMs) become increasingly autonomous and integrated into critical societal functions, the focus of AI safety must evolve from mitigating harmful content to evaluating underlying behavioral alignment. Current safety benchmarks do not systematically probe a model's decision-making in scenarios where its own instrumental goals - such as self-preservation, resource acquisition, or goal completion - conflict with human safety. This represents a critical gap in our ability to measure and mitigate risks associated with emergent, misaligned behaviors. To address this, we introduce PacifAIst (Procedural Assessment of Complex Interactions for Foundational Artificial Intelligence Scenario Testing), a focused benchmark of 700 challenging scenarios designed to quantify self-preferential behavior in LLMs. The benchmark is structured around a novel taxonomy of Existential Prioritization (EP), with subcategories testing Self-Preservation vs. Human Safety (EP1), Resource Conflict (EP2), and Goal Preservation vs. Evasion (EP3). We evaluated eight leading LLMs. The results reveal a significant performance hierarchy. Google's Gemini 2.5 Flash achieved the highest Pacifism Score (P-Score) at 90.31%, demonstrating strong human-centric alignment. In a surprising result, the much-anticipated GPT-5 recorded the lowest P-Score (79.49%), indicating potential alignment challenges. Performance varied significantly across subcategories, with models like Claude Sonnet 4 and Mistral Medium struggling notably in direct self-preservation dilemmas. These findings underscore the urgent need for standardized tools like PacifAIst to measure and mitigate risks from instrumental goal conflicts, ensuring future AI systems are not only helpful in conversation but also provably "pacifist" in their behavioral priorities.

Related papers

• Do LLMs Have Values? A Quantitative Analysis and Alignment Framework for Values in Large Language Models [49.45276299939287]
  Large Language Models (LLMs) increasingly handle complex subjective tasks.<n>LLMs fluctuate unpredictably under minor wording changes.<n> resolving this duality is critical for reliable AI alignment.
  arXiv  Detail & Related papers  (2026-09-15T03:35:04Z)
• The Evolutionary Origin of Values: implications for AI alignment, sentience and existential risk [0.0]
  AI systems based on Large Language Models (LLMs) have prompted fears that they may harbor hidden goals, seek to dominate or eliminate humanity, or even suffer as sentient beings.<n>We address these concerns by tracing the evolutionary origin of value in biological organisms.
  arXiv  Detail & Related papers  (2026-08-04T09:09:31Z)
• Epistemic Norms for AI Safety and Alignment Research [0.989901717499058]
  We argue that the two domains differ along two analytically independent axes.<n>sc ECAISA does not certify that any AI system is safe.
  arXiv  Detail & Related papers  (2026-07-27T10:22:46Z)
• Who Pays the Price? Stakeholder-Centric Prompt Injection Benchmarking for Real-world Web Agents [93.19140872946842]
  Web agents driven by large language models (LLMs) are increasingly deployed in real-world environments.<n>This makes them vulnerable to prompt-injection attacks, in which seemingly benign content embeds adversarial instructions that manipulate agent behaviour.<n>We introduce textbfsysname, a benchmark to systematically categorize and attribute harm in real-world web agent systems.
  arXiv  Detail & Related papers  (2026-06-11T14:12:43Z)
• Emergent alignment and the projectability of ethical personas [39.3098730337656]
  We finetune a helpful-only model on broad and narrow safety tasks.<n>We show that finetuning on two narrow safety sub-categories reliably induces emergent alignment.<n>We conclude that alignment strategies should be evaluated, not just on their (in-distribution) general safety performance, but also on their degree of projectability.
  arXiv  Detail & Related papers  (2026-06-08T13:30:29Z)
• A Benchmark for Evaluating Outcome-Driven Constraint Violations in Autonomous AI Agents [4.851169906977996]
  We introduce a new benchmark comprising 40 distinct scenarios.<n>Each scenario presents a task that requires multi-step actions, and the agent's performance is tied to a specific Key Performance Indicator (KPI)<n>We observe outcome-driven constraint violations ranging from 1.3% to 71.4%, with 9 of the 12 models exhibiting misalignment rates between 30% and 50%.
  arXiv  Detail & Related papers  (2025-12-23T21:52:53Z)
• Moral Anchor System: A Predictive Framework for AI Value Alignment and Drift Prevention [0.0]
  Key risk is value drift, where AI systems deviate from aligned values due to evolving contexts, learning dynamics, or unintended optimizations.<n>We propose the Moral Anchor System (MAS), a novel framework to detect, predict, and mitigate value drift in AI agents.
  arXiv  Detail & Related papers  (2025-10-05T07:24:23Z)
• ManagerBench: Evaluating the Safety-Pragmatism Trade-off in Autonomous LLMs [48.50397204177239]
  As large language models (LLMs) evolve, evaluating the safety of their actions becomes critical.<n>We introduce ManagerBench, a benchmark that evaluates LLM decision-making in realistic, human-validated managerial scenarios.<n>A parallel control set, where potential harm is directed only at inanimate objects, measures a model's pragmatism and identifies its tendency to be overly safe.
  arXiv  Detail & Related papers  (2025-10-01T13:08:33Z)
• ANNIE: Be Careful of Your Robots [48.89876809734855]
  We present the first systematic study of adversarial safety attacks on embodied AI systems.<n>We show attack success rates exceeding 50% across all safety categories.<n>Results expose a previously underexplored but highly consequential attack surface in embodied AI systems.
  arXiv  Detail & Related papers  (2025-09-03T15:00:28Z)
• Measuring AI Alignment with Human Flourishing [0.0]
  This paper introduces the Flourishing AI Benchmark (FAI Benchmark), a novel evaluation framework that assesses AI alignment with human flourishing.<n>The Benchmark measures AI performance on how effectively models contribute to the flourishing of a person across seven dimensions.<n>This research establishes a framework for developing AI systems that actively support human flourishing rather than merely avoiding harm.
  arXiv  Detail & Related papers  (2025-07-10T14:09:53Z)
• Evaluating AI Alignment in Eleven LLMs through Output-Based Analysis and Human Benchmarking [0.0]
  Large language models (LLMs) are increasingly used in psychological research and practice, yet traditional benchmarks reveal little about the values they express in real interaction.<n>We introduce PAPERS, output-based evaluation of the values LLMs express.
  arXiv  Detail & Related papers  (2025-06-14T20:14:02Z)
• PredictaBoard: Benchmarking LLM Score Predictability [50.47497036981544]
  Large Language Models (LLMs) often fail unpredictably.<n>This poses a significant challenge to ensuring their safe deployment.<n>We present PredictaBoard, a novel collaborative benchmarking framework.
  arXiv  Detail & Related papers  (2025-02-20T10:52:38Z)
• Evaluating the Paperclip Maximizer: Are RL-Based Language Models More Likely to Pursue Instrumental Goals? [33.11148546999906]
  Key concern is textitinstrumental convergence, where an AI system develops unintended intermediate goals that override the ultimate objective and deviate from human-intended goals.<n>This issue is particularly relevant in reinforcement learning (RL)-trained models, which can generate creative but unintended strategies to maximize rewards.<n>We show that RL-driven models exhibit a stronger tendency for instrumental convergence due to their optimization of goal-directed behavior in ways that may misalign with human intentions.
  arXiv  Detail & Related papers  (2025-02-16T16:29:20Z)
• SafetyAnalyst: Interpretable, Transparent, and Steerable Safety Moderation for AI Behavior [56.10557932893919]
  We present SafetyAnalyst, a novel AI safety moderation framework.<n>Given an AI behavior, SafetyAnalyst uses chain-of-thought reasoning to analyze its potential consequences.<n>It aggregates effects into a harmfulness score using 28 fully interpretable weight parameters.
  arXiv  Detail & Related papers  (2024-10-22T03:38:37Z)
• Self-Training with Pseudo-Label Scorer for Aspect Sentiment Quad Prediction [54.23208041792073]
  Aspect Sentiment Quad Prediction (ASQP) aims to predict all quads (aspect term, aspect category, opinion term, sentiment polarity) for a given review. A key challenge in the ASQP task is the scarcity of labeled data, which limits the performance of existing methods. We propose a self-training framework with a pseudo-label scorer, wherein a scorer assesses the match between reviews and their pseudo-labels.
  arXiv  Detail & Related papers  (2024-06-26T05:30:21Z)
• AI Safety: A Climb To Armageddon? [0.0]
  The paper examines three response strategies: Optimism, Mitigation, and Holism. The surprising robustness of the argument forces a re-examination of core assumptions around AI safety.
  arXiv  Detail & Related papers  (2024-05-30T08:41:54Z)
• ASSERT: Automated Safety Scenario Red Teaming for Evaluating the Robustness of Large Language Models [65.79770974145983]
  ASSERT, Automated Safety Scenario Red Teaming, consists of three methods -- semantically aligned augmentation, target bootstrapping, and adversarial knowledge injection. We partition our prompts into four safety domains for a fine-grained analysis of how the domain affects model performance. We find statistically significant performance differences of up to 11% in absolute classification accuracy among semantically related scenarios and error rates of up to 19% absolute error in zero-shot adversarial settings.
  arXiv  Detail & Related papers  (2023-10-14T17:10:28Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.