# Secure Code Benchmark Finds Top Coding Agents Write Secure Code Only 25% of the Time

DevFeed: [Secure Code Benchmark Finds Top Coding Agents Write Secure Code Only 25% of the Time](<https://devfeed.tech/articles/secure-code-benchmark-finds-top-coding-agents-write-secure-code-only-25-of-the-time-122643.md>)

Original publisher: [Read original article](<https://hackernoon.com/secure-code-benchmark-finds-top-coding-agents-write-secure-code-only-25percent-of-the-time?source=rss>)

Author: Qwen AI (Alibaba)

Published: 2026-11-08T13:21:08Z

Content type: news

Language: en

Sources: [HackerNoon](<https://devfeed.tech/sources/hackernoon.md>)

Topics: [code security](<https://devfeed.tech/topics/code-security.md>), [agentic-coding](<https://devfeed.tech/topics/agentic-coding.md>), [Benchmark](<https://devfeed.tech/topics/benchmark.md>), [AI-assisted coding](<https://devfeed.tech/topics/ai-assisted-coding.md>), [Spec Driven Development](<https://devfeed.tech/topics/spec-driven-development.md>), [Claude Code](<https://devfeed.tech/topics/claude-code.md>), [codex](<https://devfeed.tech/topics/codex.md>), [Static code analysis](<https://devfeed.tech/topics/static-code-analysis.md>)

Tags: [2](<https://devfeed.tech/tags/2.md>), [2021](<https://devfeed.tech/tags/2021.md>), [2022](<https://devfeed.tech/tags/2022.md>), [agent](<https://devfeed.tech/tags/agent.md>), [agents](<https://devfeed.tech/tags/agents.md>), [ai](<https://devfeed.tech/tags/ai.md>), [ai-coding](<https://devfeed.tech/tags/ai-coding.md>), [ai-coding-agent](<https://devfeed.tech/tags/ai-coding-agent.md>), [ai-coding-agents](<https://devfeed.tech/tags/ai-coding-agents.md>), [benchmark](<https://devfeed.tech/tags/benchmark.md>), [best](<https://devfeed.tech/tags/best.md>), [code](<https://devfeed.tech/tags/code.md>), [coding](<https://devfeed.tech/tags/coding.md>), [coding-agent](<https://devfeed.tech/tags/coding-agent.md>), [coding-agents](<https://devfeed.tech/tags/coding-agents.md>), [coding-agents-code-security](<https://devfeed.tech/tags/coding-agents-code-security.md>), [cves](<https://devfeed.tech/tags/cves.md>), [cybersecurity](<https://devfeed.tech/tags/cybersecurity.md>), [hackernoon-top-story](<https://devfeed.tech/tags/hackernoon-top-story.md>), [secure-code-benchmark](<https://devfeed.tech/tags/secure-code-benchmark.md>), [secure-coding](<https://devfeed.tech/tags/secure-coding.md>), [shows](<https://devfeed.tech/tags/shows.md>), [time](<https://devfeed.tech/tags/time.md>), [top](<https://devfeed.tech/tags/top.md>), [vibe-coding](<https://devfeed.tech/tags/vibe-coding.md>), [write](<https://devfeed.tech/tags/write.md>)

## AI overview

LSCBench evaluates whether coding agents produce code that is both functional and secure across 450 tasks derived from 150 real vulnerability scenarios, seven languages, and function, file, and repository scopes. Tasks specify functionality without mentioning security, and hidden tests separately assess functional behavior and security. Across six frontier agents, functional pass rates were 73.6%-91.3%, but the best joint pass rate was 25.1%; the best agent secured all three scopes for only about 12% of scenarios. A reminder to write secure code raised joint pass rates by 15-21 points for three of four rerun models, with little change in functional success, yet most tasks remained insecure. The reported results suggest agents more often handle local checks such as bounds checks than defenses requiring unstated policies, such as access or data disclosure rules. The authors note that results concern one benchmark and its task setup.

## Source excerpt

A 450-task benchmark built from real CVEs shows even the best AI coding agent writes code that's both working and secure only 25% of the time.