Failure Analysis
Why System Prompts Leak: Failure Modes and Safer Design
Discover how to prevent system prompt leaks, a critical risk in LLM design. Learn best practices to enhance security and mitigate exposure.
11 MIN READ
The AI Failure Library
Blog
Analysis and engineering practice from the AI Failure Library, written against the cases in the record rather than around them.
Recent articles
AI Risk Fundamentals
Dataset Shift Explained: How Production Data Breaks ML Assumptions
Discover how dataset shift impacts machine learning models in production and learn effective strategies to address this challenge.
14 MIN READ
Failure by Industry
Customer-Service AI Hallucinations: Failure Patterns, Liability, and Guardrails
Explore crucial model hallucination examples that highlight risks in customer-service AI, revealing how false information can have serious consequences.
16 MIN READ
AI Risk Fundamentals
LLM Jailbreaks vs Prompt Injection: The Difference and Why It Matters
Discover effective strategies to test and defend against LLM jailbreaks, ensuring your AI models adhere to safety and compliance standards.
18 MIN READ
AI Risk Fundamentals
Reward Hacking Explained: When AI Optimizes the Wrong Goal
Discover how to tackle reward hacking in AI systems. Learn to audit your reward signals and prevent common pitfalls in design.
20 MIN READ
Tools & Evaluation
AI Audit Checklist: Evidence to Review Across the System Lifecycle
Expand your AI audit strategy with a detailed checklist that covers evidence across the entire system lifecycle to ensure compliance.
28 MIN READ
Prevention & Reliability
AI Supply-Chain Security: Models, Data, Adapters, and Dependencies
Enhance your AI supply chain security by enforcing integrity checks, cryptographic signatures, and monitoring practices to prevent attacks.
19 MIN READ
AI Risk Fundamentals
Membership Inference Attacks: What They Reveal and How to Test for Them
Discover how to test for membership inference attacks and protect sensitive data. Learn essential strategies to safeguard your models.
22 MIN READ
Failure Analysis
Machine-Learning Failure Analysis: From Symptom to Verified Cause
Discover how to effectively conduct ML failure analysis to identify root causes and implement reliable fixes, ensuring robust models.
21 MIN READ
Prevention & Reliability
ISO/IEC 42001:2023 for AI Teams - Scope, Implementation, and Limits
Discover how ISO 42001:2023 enhances AI management for teams by guiding effective implementation and continuous improvement in your organization.
16 MIN READ
Prevention & Reliability
AI Incident Response: A Practical Runbook for GenAI Systems
Master AI incident response with our practical playbook, guiding security teams through containment, evidence preservation, and swift action.
25 MIN READ
AI Risk Fundamentals
AI Risk Taxonomy: A Practical Way to Classify Failures and Harms
Learn to effectively classify AI risks with a practical taxonomy that guides your team through incident management and mitigation steps.
18 MIN READ