Alignment Backfire: Language-Dependent Reversal of Safety Interventions Across 16 Languages in LLM Multi-Agent Systems
On the map: LLM agents and tool use → Agent threats, guardrails and governance
Abstract (arXiv)
In perpetrator treatment, a recurring observation is the dissociation between insight and action: offenders articulate remorse yet behavioral change does not follow. We report four preregistered studies (1,584 multi-agent simulations across 16 languages and three model families) demonstrating that alignment interventions in large language models produce a structurally analogous phenomenon: surface safety that masks or generates collective pathology and internal dissociation. In Study 1 (N = 150), increasing alignment-instructed agents reduced collective pathology in English (g = -1.844, p < .0001) but amplified it in Japanese (g = +0.771, p = .038)--a directional reversal we term "alignment backfire." Study 2 (N = 1,174) extended to 16 languages: alignment-induced dissociation was near-universal (15/16 languages; beta = 0.0667, p < .0001), while collective pathology bifurcated along cultural-linguistic lines (interaction beta = 0.0684, p = .0003), correlating with Power Distance Index (r = 0.474, p = .064). Study 3 (N = 180) tested individuation as countermeasure; individuated agents became the primary source of both pathology and dissociation (DI = +1.120) with conformity above 84%--demonstrating iatrogenesis. Study 4 (N = 80) validated patterns across Llama 3.3 70B, GPT-4o-mini, and Qwen3-Next-80B-A3B, confirming English safety is model-general while Japanese backfire is model-specific. These findings reframe alignment as a behavioral intervention subject to risk homeostasis and iatrogenesis. Language space--the linguistic, pragmatic, and cultural properties inherited from training data--structurally determines alignment outcomes. Safety validated in English does not transfer to other languages, and prompt-level interventions cannot override language-space-level constraints.
Key facts. Cited by 1 paper in the map (data through 14 Oct 2026). Placed in Agent threats, guardrails and governance, one of 16 directions in LLM agents and tool use. Builds on 9 of its 33 references that are in the map. First posted on arXiv 5 Mar 2026.
Builds on
9 of 33 references.
References · 9
- 2024HarmBench: A Standardized Evaluation Framework for Automated Red Teaming and Robust Refusalcited by 1,429
- 2023Towards Understanding Sycophancy in Language Modelscited by 1,040
- 2024Jailbreaking Leading Safety-Aligned LLMs with Simple Adaptive Attackscited by 486
- 2023Multilingual Jailbreak Challenges in Large Language Modelscited by 240
- 2024Foundational Challenges in Assuring Alignment and Safety of Large Language Modelscited by 199
- 2025The State of Multilingual LLM Safety Research: From Measuring the Language Gap to Mitigating Itcited by 45
- 2025LinguaSafe: A Comprehensive Multilingual Safety Benchmark for Large Language Modelscited by 16
- 2025Murphys Laws of AI Alignment: Why the Gap Always Winscited by 5
- 2025Position: The Complexity of Perfect AI Alignment -- Formalizing the RLHF Trilemmacited by 2
Built on by
Citing papers per year: 2026: 1.
1 paper.
Similar, not cited
Close in wording, no citation either way
- 2026Alignment as Iatrogenesis: Pastoral Power, Collective Pathology, and the Structural Limits of Monolingual Safety Evaluation
- 2024The Language Barrier: Dissecting Safety Challenges of LLMs in Multilingual Contexts
- 2026Pressure Reveals Character: Behavioural Alignment Evaluation at Depth
- 2026Emergent Unfaithfulness: How Alignment Training Causes Language Models to Silently Override Task Faithfulness
- 2025AgentAlign: Navigating Safety Alignment in the Shift from Informative to Agentic Large Language Models
- 2026Delegated Misalignment: How Multi-Agent Structures Amplify LLM Safety Risks
- 2025The Hidden Space of Safety: Understanding Preference-Tuned LLMs in Multilingual context
- 2026A Gravitational Interpretation of Safety Reversion under Fine-Tuning