PulseAugur
EN
LIVE 13:08:45
Deutsch(DE) Seit zwei Jahren hängt die Security-Pass-Rate von LLM-Code bei ~55 % fest. GPT-5, Gemini 3, Claude 4, ... In fast jeder zweiten Aufgabe baut das Modell eine bek

LLM Code Security Pass Rate Stagnates at 55% Despite New Models

By PulseAugur Editorial · [1 sources] ·

Despite advancements in models like GPT-5.5, Gemini 3, and Claude 4, the security pass rate for LLM-generated code has remained stagnant at approximately 55% for two years. These models frequently introduce known security vulnerabilities in nearly half of the tasks they handle, even though their syntactic correctness is high. While LLMs can increase coding speed, they do not inherently improve the security of delivered software. AI

IMPACT LLM-generated code continues to introduce security vulnerabilities, indicating a need for improved security practices and tools beyond simple code generation.

RANK_REASON The item discusses a research finding about the security pass rate of LLM-generated code, citing a specific benchmark and mentioning multiple LLM models. [lever_c_demoted from research: ic=1 ai=1.0]

Read on Mastodon — fosstodon.org →

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →

LLM Code Security Pass Rate Stagnates at 55% Despite New Models

How we ranked this

Signal score
0 / 100
Composite score across the factors below. Higher = stronger signal that this story matters right now.
Newsworthiness bucket
Tool
The item discusses a research finding about the security pass rate of LLM-generated code, citing a specific benchmark and mentioning multiple LLM models. [lever_c_demoted from research: ic=1 ai=1.0]
Source corroboration
Single-source cluster
Only one publisher covered this so far. Single-source stories can still rank when the publisher is high-authority, but they lack cross-source corroboration.
Topics
safety, model release
Editorial topic classification. Feeds into how the story surfaces on /topic/<slug> hub pages and into the per-entity coverage mix.
AI-industry relevance
High
Clearly on-topic for AI-industry coverage.
Story freshness
97 days old
Aged out of breaking-news scoring windows; ranking reflects the durable signal from the full source set.

Full methodology in our editorial standards.

COVERAGE [1]

  1. Mastodon — fosstodon.org TIER_1 Deutsch(DE) · [email protected] ·

    For two years, the security pass rate of LLM code has been stuck at ~55%. GPT-5.5, Gemini 3, Claude 4, ... In almost every second task, the model builds a vulnerability

    Seit zwei Jahren hängt die Security-Pass-Rate von LLM-Code bei ~55 % fest. GPT-5, Gemini 3, Claude 4, ... In fast jeder zweiten Aufgabe baut das Modell eine bekannte Sicherheitslücke ein. Syntaktisch sind sie quasi perfekt (>95 %). Sicher werden sie nicht. LLMs macht dein Team sc…

    LINKS veracode.com/…/spring-2026-genai-code-sec…

RELATED ENTITIES

RELATED TOPICS