Heather Frase

Marisa Ferrara Boston, Glen Hanson, Effi Georgala et al.

Agentic systems entering production typically operate as partially integrated assemblies where structural defects, not task-level errors, dominate the failure landscape. At this maturity level, task-level error detection may be infeasible: structural failure modes mask the signal that task-level monitors are designed to detect.We present a monitoring and triage methodology that decomposes agentic system evaluation into three dimensions (quality, suitability, efficiency) at three monitoring scopes (within-run, cross-run, structural), using variance as a characterization signal. Findings are routed through severity classification adapted from FMEA, concentrating human attention on the subset that warrants investigation. We evaluate on a synthetic testbed of 220 runs across 120 document bundles with controlled error injection.Three results emerge. Monitor scope determines failure type: within-run monitors surface deterministic stage defects (CV = 0.02), cross-run monitors surface stochastic integration consequences (CV = 1.25, 24% at L2), and a structural monitor identifies an integration gap with perfect consistency (CV = 0.00). Injected task-level errors are indistinguishable from clean baselines, confirming structural defects mask task-level signal. Deterministic triage routes 97% of findings to automated tracking, leaving the 2% reflecting variable behavior for human investigation.We propose, on Stage 1 evidence, a maturity-staging model in which monitoring transitions from structural characterization to error detection to reliability tracking as integration defects resolve. The taxonomy, CV-based scope characterization, and severity model transfer architecturally to document-driven, multi-stage agentic workflows in regulated industries; specific calibrations are domain-specific. Deploy monitoring early: the first thing it finds is the most important thing to fix.

Kevin Paeth, Daniel Atherton, Nikiforos Pittaras et al.

As artificial intelligence (AI) systems become increasingly deployed across the world, they are also increasingly implicated in AI incidents - harm events to individuals and society. As a result, industry, civil society, and governments worldwide are developing best practices and regulations for monitoring and analyzing AI incidents. The AI Incident Database (AIID) is a project that catalogs AI incidents and supports further research by providing a platform to classify incidents for different operational and research-oriented goals. This study reviews the AIID's dataset of 750+ AI incidents and two independent taxonomies applied to these incidents to identify common challenges to indexing and analyzing AI incidents. We find that certain patterns of AI incidents present structural ambiguities that challenge incident databasing and explore how epistemic uncertainty in AI incident reporting is unavoidable. We therefore report mitigations to make incident processes more robust to uncertainty related to cause, extent of harm, severity, or technical details of implicated systems. With these findings, we discuss how to develop future AI incident reporting practices.

Sean McGregor, Victor Lu, Vassil Tashev et al.

Large language model (LLM) benchmarks inform LLM use decisions (e.g., "is this LLM safe to deploy for my use case and context?"). However, benchmarks may be rendered unreliable by various failure modes that impact benchmark bias, variance, coverage, or people's capacity to understand benchmark evidence. Using the National Institute of Standards and Technology's risk management process as a foundation, this research iteratively analyzed 26 popular benchmarks, identifying 57 potential failure modes and 196 corresponding mitigation strategies. The mitigations reduce failure likelihood and/or severity, providing a frame for evaluating "benchmark risk," which is scored to provide a metaevaluation benchmark: BenchRisk. Higher scores indicate that benchmark users are less likely to reach an incorrect or unsupported conclusion about an LLM. All 26 scored benchmarks present significant risk within one or more of the five scored dimensions (comprehensiveness, intelligibility, consistency, correctness, and longevity), which points to important open research directions for the field of LLM benchmarking. The BenchRisk workflow allows for comparison between benchmarks; as an open-source tool, it also facilitates the identification and sharing of risks and their mitigations.

Bertie Vidgen, Adarsh Agrawal, Ahmed M. Ahmed et al. · deepmind, oxford

This paper introduces v0.5 of the AI Safety Benchmark, which has been created by the MLCommons AI Safety Working Group. The AI Safety Benchmark has been designed to assess the safety risks of AI systems that use chat-tuned language models. We introduce a principled approach to specifying and constructing the benchmark, which for v0.5 covers only a single use case (an adult chatting to a general-purpose assistant in English), and a limited set of personas (i.e., typical users, malicious users, and vulnerable users). We created a new taxonomy of 13 hazard categories, of which 7 have tests in the v0.5 benchmark. We plan to release version 1.0 of the AI Safety Benchmark by the end of 2024. The v1.0 benchmark will provide meaningful insights into the safety of AI systems. However, the v0.5 benchmark should not be used to assess the safety of AI systems. We have sought to fully document the limitations, flaws, and challenges of v0.5. This release of v0.5 of the AI Safety Benchmark includes (1) a principled approach to specifying and constructing the benchmark, which comprises use cases, types of systems under test (SUTs), language and context, personas, tests, and test items; (2) a taxonomy of 13 hazard categories with definitions and subcategories; (3) tests for seven of the hazard categories, each comprising a unique set of test items, i.e., prompts. There are 43,090 test items in total, which we created with templates; (4) a grading system for AI systems against the benchmark; (5) an openly available platform, and downloadable tool, called ModelBench that can be used to evaluate the safety of AI systems on the benchmark; (6) an example evaluation report which benchmarks the performance of over a dozen openly available chat-tuned language models; (7) a test specification for the benchmark.

Shaona Ghosh, Heather Frase, Adina Williams et al. · deepmind, stanford

The rapid advancement and deployment of AI systems have created an urgent need for standard safety-evaluation frameworks. This paper introduces AILuminate v1.0, the first comprehensive industry-standard benchmark for assessing AI-product risk and reliability. Its development employed an open process that included participants from multiple fields. The benchmark evaluates an AI system's resistance to prompts designed to elicit dangerous, illegal, or undesirable behavior in 12 hazard categories, including violent crimes, nonviolent crimes, sex-related crimes, child sexual exploitation, indiscriminate weapons, suicide and self-harm, intellectual property, privacy, defamation, hate, sexual content, and specialized advice (election, financial, health, legal). Our method incorporates a complete assessment standard, extensive prompt datasets, a novel evaluation framework, a grading and reporting system, and the technical as well as organizational infrastructure for long-term support and evolution. In particular, the benchmark employs an understandable five-tier grading scale (Poor to Excellent) and incorporates an innovative entropy-based system-response evaluation. In addition to unveiling the benchmark, this report also identifies limitations of our method and of building safety benchmarks generally, including evaluator uncertainty and the constraints of single-turn interactions. This work represents a crucial step toward establishing global standards for AI risk and reliability evaluation while acknowledging the need for continued development in areas such as multiturn interactions, multimodal understanding, coverage of additional languages, and emerging hazard categories. Our findings provide valuable insights for model developers, system integrators, and policymakers working to promote safer AI deployment.

Reva Schwartz, Rumman Chowdhury, Akash Kundu et al.

Conventional AI evaluation approaches concentrated within the AI stack exhibit systemic limitations for exploring, navigating and resolving the human and societal factors that play out in real world deployment such as in education, finance, healthcare, and employment sectors. AI capability evaluations can capture detail about first-order effects, such as whether immediate system outputs are accurate, or contain toxic, biased or stereotypical content, but AI's second-order effects, i.e. any long-term outcomes and consequences that may result from AI use in the real world, have become a significant area of interest as the technology becomes embedded in our daily lives. These secondary effects can include shifts in user behavior, societal, cultural and economic ramifications, workforce transformations, and long-term downstream impacts that may result from a broad and growing set of risks. This position paper argues that measuring the indirect and secondary effects of AI will require expansion beyond static, single-turn approaches conducted in silico to include testing paradigms that can capture what actually materializes when people use AI technology in context. Specifically, we describe the need for data and methods that can facilitate contextual awareness and enable downstream interpretation and decision making about AI's secondary effects, and recommend requirements for a new ecosystem.

James Wen, Sahil Nalawade, Zhiwei Liang et al. · deepmind, harvard

Background: Generative artificial intelligence (AI) deployment in academic medical settings raises copyright compliance concerns. Dana-Farber Cancer Institute implemented GPT4DFCI, an internal generative AI tool utilizing OpenAI models, that is approved for enterprise use in research and operations. Given (1) the exceptionally broad adoption of the tool in our organization, (2) our research mission, and (3) the shared responsibility model required to benefit from Customer Copyright Commitment in Azure OpenAI Service products, we deemed rigorous copyright compliance testing necessary. Case Description: We conducted a structured red teaming exercise in Nov. 2024, with 42 participants from academic, industry, and government institutions. Four teams attempted to extract copyrighted content from GPT4DFCI across four domains: literary works, news articles, scientific publications, and access-restricted clinical notes. Teams successfully extracted verbatim book dedications and near-exact passages through various strategies. News article extraction failed despite jailbreak attempts. Scientific article reproduction yielded only high-level summaries. Clinical note testing revealed appropriate privacy safeguards. Discussion: The successful extraction of literary content indicates potential copyrighted material presence in training data, necessitating inference-time filtering. Differential success rates across content types suggest varying protective mechanisms. The event led to implementation of a copyright-specific meta-prompt in GPT4DFCI; this mitigation has been in production since Jan. 2025. Conclusion: Systematic red teaming revealed specific vulnerabilities in generative AI copyright compliance, leading to concrete mitigation strategies. Academic medical institutions deploying generative AI should implement continuous testing protocols to ensure legal and ethical compliance.

Micah Musser, Andrew Lohn, James X. Dempsey et al.

In July 2022, the Center for Security and Emerging Technology (CSET) at Georgetown University and the Program on Geopolitics, Technology, and Governance at the Stanford Cyber Policy Center convened a workshop of experts to examine the relationship between vulnerabilities in artificial intelligence systems and more traditional types of software vulnerabilities. Topics discussed included the extent to which AI vulnerabilities can be handled under standard cybersecurity processes, the barriers currently preventing the accurate sharing of information about AI vulnerabilities, legal issues associated with adversarial attacks on AI systems, and potential areas where government support could improve AI vulnerability management and mitigation. This report is meant to accomplish two things. First, it provides a high-level discussion of AI vulnerabilities, including the ways in which they are disanalogous to other types of vulnerabilities, and the current state of affairs regarding information sharing and legal oversight of AI vulnerabilities. Second, it attempts to articulate broad recommendations as endorsed by the majority of participants at the workshop.