Fuzz testing
Also written: Fuzzing
Automated software testing that feeds random or unexpected data to a program to find bugs.
Fuzz testing, or fuzzing, is a way to find bugs in software. Testers use tools to automatically send lots of random or malformed data to a program. The goal is to see if the program crashes, freezes, or behaves in unexpected ways when it encounters this bad input.
Think of it like shaking a vending machine to see if you can get free snacks. Fuzzing does something similar for code. If the program handles the unexpected input gracefully, it's likely more robust. If it breaks, the developers know where to fix it.
Fuzz testing is an automated software testing technique that involves providing invalid, unexpected, or random data as input to a computer program. The primary goal is to discover software defects, such as crashes, memory leaks, or security vulnerabilities, by triggering edge cases or unhandled conditions. Fuzzing can be applied at various levels, from individual functions to entire applications, and often involves mutation-based fuzzing (modifying valid inputs) or generation-based fuzzing (creating inputs from scratch based on a model).
Frontier AI labs are increasingly using fuzz testing to ensure the reliability and safety of their complex AI systems. As these systems become more integrated into software development, finding bugs early through methods like fuzzing is crucial for maintaining code quality and preventing unexpected behavior.
- Finding bugs with Claude and property-based testingAnthropic · Anthropic Research ·
- Anthropic Claude
- Anthropic's AI agents are being used to perform fuzz testing, demonstrating how AI can assist in finding bugs in software.
- Unit testing
- Unit tests focus on verifying small, isolated pieces of code with
- Integration testing
- Integration tests check how different software modules work
- Static analysis
- Static analysis examines code without executing it to find pote
Fuzz testing originated in the 1980s with Barton Miller at the University of Wisconsin. He developed techniques to test the robustness of Unix utilities by feeding them random input. Source ↗
While direct fuzz testing is complex, you can observe related concepts. Try inputting unusual characters or very long strings into a text field on a website or app. See if the application handles it gracefully or if it displays an error or unexpected behavior. This mimics the basic idea of feeding unexpected data.
The Playbook: reusable how-tos, prompts, configs and tips pulled out of the news.
- PromptRole-play a student struggling with math for AI tutor simulationSimulate a student's math struggles and uncertainty for AI tutor evaluation. · 5 min
- TemplateReusable workflow templates for healthcare teamsTeams can create and share templates for common workflows like documentation or prior authorizations. · 15 min
- TemplateTemplate for building an AI tool tailored to your workA structured approach to defining and building a custom AI tool for specific business needs. · 15 min
- TemplateEvals CSV prompt list for agent skillsA CSV file template for listing prompts used in evaluating agent skills. · 20 min
- HeuristicDouble coding agent task completion with docs linksProviding coding agents with a documentation link significantly increases their task completion rate. · 2 min
Questions people ask
- What is the main goal of fuzz testing?
- The main goal is to discover software bugs, crashes, and security vulnerabilities by feeding a program unexpected or random data.
- Is fuzz testing only for security?
- No, while it's excellent for finding security flaws, fuzz testing also helps improve general software stability and robustness by uncovering unexpected behavior.
First seen on Subvolts Jun 17, 2026. Updated Sep 14, 2026. Written by an AI model from the items above and reviewed against them weekly; report an error.