Overview

The field asks how to make AI systems behave reliably, remain controllable, and avoid harms to people and society. Approaches include alignment training, red-teaming and benchmarks, interpretability, formal verification, runtime monitoring, and safety engineering for systems that use tools or act in the physical world. Researchers also study risks in deployment, including misuse, overreliance, organizational failures, and effects on vulnerable users, using methods from human-computer interaction, law, and governance. A central open tension is that strong results on bounded tests do not establish safety in long, changing real-world settings, while comprehensive guarantees for complex systems remain difficult to obtain.

Latest papers

CardsList

143 papers15 papers in the last four weeks, down 12% on the four weeks before. 0.1% of all new papers.

  1. Designing Deceptive Model Organisms that Successfully Evade Auditing

    Nov 16, 2026Vasisht Duddu, Mark Wang, Lipeng He +1AI SafetyDeception in Language Models

  2. Between Rigor and Reality: How AI Safety Benchmark Developers Understand Benchmark Use and Maintenance

    Nov 13, 2026Megan Li, Ningjing Tang, Christopher Stewart +2AI Safety EvaluationAI Safety

  3. AI Alignment Is Neither Solvable nor Sufficient

    Nov 9, 2026Gerrit GroßmannAI AlignmentAI Safety

  4. Open Problems in Chatbots and Mental Health

    Nov 7, 2026Jared Moore, Jacy Reese Anthis, William Agnew +8Mental Health Conversational AgentsAI Safety

  5. AYO: A Safe and Privacy-Preserving Platform for Studying Children's Interaction with LLMs

    Nov 5, 2026Huiwen Alex Yang, Rachel Skwersky, Bill D. Thompson +1Human-AI InteractionAI Safety

  6. Understanding AI from the Inside: The State of Mechanistic Interpretability for AI Safety

    Nov 5, 2026Fazl Barez, Tung-Yu Wu, Hunar Batra +24Mechanistic InterpretabilityAI Safety

  7. Adapting the Process of Adaptation: Darwinian, Lamarckian, and Agent-Directed Evolution in AI

    Nov 4, 2026Andrew W. Lo, Chaoyi ZhaoSelf-Evolving AgentsAI Safety

  8. A Definition of Catastrophic Performance

    Nov 4, 2026Henrik Marklund, Alex Infanger, Benjamin Van RoyAI AlignmentAI Safety

  9. Safe by Benchmark, Unsafe in Practice: Lessons from 1331 Child-Related AI Incidents

    Oct 26, 2026Junfeng Jiao, Kevin Chen, Shrestha Mishra +4AI SafetyAI Accountability

  10. From Clinical Judgment to Checkable Rules: Clinician-Led Guardrails for Mental-Health Chatbots

    Oct 22, 2026Chenhan Lyu, Kit Wislocki, Sarah Elizabeth Stoeckl +2LLM GuardrailsLLM Safety

  11. COMET: A coordinated multi-layer runtime monitoring framework for human-supervised AI-enabled safety systems

    Oct 21, 2026Niclas Flehmig, Mary Ann Lundteigen, Shen YinAI SafetyAI Agent Monitoring

  12. Humans Correct Their Judgments Through Debate, but Weaker AI Models Do Not

    Oct 20, 2026Maria Victoria Carro, Denise Alejandra Mester, Facundo Nieto +14Multi-Agent DebateHuman Oversight of AI

  13. Fundamental Impossibility Results in AI Safety: Trading Off between Desiderata

    Oct 19, 2026Ida Mattsson, Ratip Emin Berker, Emanuel Tewolde +11AI SafetyInformation Design

  14. Towards Robust Model Organisms for Eliciting Latent Knowledge

    Oct 19, 2026Lennart Finke, Hardik Bhatnagar, Euodia Dodd +4AI SafetyLanguage Model Safety Evaluation

  15. FinSafeGuard: A Quality-Driven Synthetic Dataset for Financial AI Safety

    Oct 19, 2026Joseph Thomas Thacil, Reetu Raj Harsh, Krati Saxena +2Synthetic Data GenerationSynthetic Data Evaluation

  16. Assessing the Empirical Evidence for Loss of Control from Agentic General-Purpose AI

    Oct 18, 2026Risto Uuk, Santeri Koivula, Lorenzo Pacchiardi +3AI Agent SafetyAI Safety

  17. AI Rights with an Off-Switch: Rights-Balancing and Due Process Harmonize AI Rights with AI Safety

    Oct 17, 2026Heather Alexander, Jonathan SimonAI EthicsAI Governance

  18. What Social Media Teaches Us About Accumulative AI Risks

    Oct 15, 2026Dawei Li, Bohan Jiang, Huan LiuAI SafetyLLM Safety

  19. Market Incentives for AI Safety Investment

    Oct 15, 2026Nikita TsoyAI SafetyAI Governance

  20. CompanionSafe: From Multi-turn Safety Evaluation to Guardrails for AI Companions

    Oct 13, 2026Xinyang Liao, Lingyu Li, Kaicheng Shen +4AI SafetyAI Safety Evaluation

  21. Superintelligence and the Triple Risk of an AI Race

    Oct 12, 2026Matthew Rendall, Max Hellrigel-Holderbaum, Leonard DungAI GovernanceArtificial General Intelligence

  22. Beyond the Ergodic Wall: A Discrete Geometric Physics Sandbox for Analysing AI Scaling Limits and Complexity Collapse

    Oct 7, 2026Simon Richard DanielAI Safety

  23. AI Safety Considerations for Agents With Limited Time to Act

    Oct 7, 2026Leo Zeitler, Jack Richings, Victoria NocklesAI Agent SafetyAI Safety

  24. Comprehension Audits to Mitigate Risks from Automated AI Research

    Oct 7, 2026Ronald J. Bodkin, Bahrad A. Sokhansanj, Gillian K. HadfieldHuman Oversight of AIAI Safety

  25. Systemization of Knowledge (SoK): Human-Centered AI Safety for Youth

    Oct 6, 2026Pratyasha Saha, Yaman Yu, Yang WangAI Risk ManagementAI Safety Evaluation

  26. How Could AI Eliminate Humanity? A Failure-Mode Analysis of Civilizational Risk

    Oct 6, 2026Mikołaj Sienicki, Krzysztof SienickiAI SafetyAI Risk Management

  27. Not What a Child Expressed: Auditing the Sign-to-Text Safety Interface in Child-Facing AI

    Oct 5, 2026Muhammad Rafiullah Memon, Viet Vo, Wanlun Ma +1Sign Language TranslationAI Safety Evaluation