Skip to main content
Mallory
← Back to News
ai-platform-security

AI Coding Assistants' Impact on Software Security and Quality

Emerging Technology Risk · 3 sources · Updated 5mo ago · First seen Nov 19, 2025

AI Coding Assistants' Impact on Software Security and Quality

OpenAI's GPT-5 and GPT-5-mini models have demonstrated significant improvements in generating secure code compared to previous iterations, according to Veracode's GenAI Code Security Report. These reasoning models achieved security pass rates of 70% and 72% on a benchmark covering common vulnerabilities such as SQL injection, weak encryption, cross-site scripting, and log injection across multiple programming languages. Despite these gains, the report notes that even the best-performing models still make insecure coding choices about 30% of the time, and non-reasoning models like GPT-5-chat lag behind, highlighting the importance of reasoning steps in AI-generated code security.

While AI coding assistants can potentially enhance software security and quality, current deployment practices often result in the rapid introduction of new vulnerabilities and quality issues. Industry experts warn that the increased speed of code generation enabled by AI is exacerbating an ongoing decline in software quality, making it harder for organizations to manage and remediate bugs. The combination of faster development cycles and insufficient oversight may lead to a net-negative impact on business security and software reliability, underscoring the need for improved development models and rigorous code review processes when leveraging AI tools.

New surface, new adversariesTrack how attackers are adapting to this technology.
Track this risk
Event timeline

How this story unfolded

Events from the most recent confirmed update back to the earliest known activity.

5 events · May 8, 2023 – Nov 19, 2025
  1. Nov 19, 20251y ago

    SC Media reports GPT-5 security benchmark results

    SC Media reported on Veracode's findings that GPT-5 reasoning models generated more secure code than prior models, while emphasizing that even the best models still made insecure choices about 30% of the time and require human review and layered security controls.

  2. Nov 18, 20251y ago

    ReversingLabs publishes analysis on AI-driven software quality decline

    ReversingLabs published a blog post arguing that AI is accelerating a broader collapse in software quality. No additional event details were provided in the reference content.

  3. Oct 1, 20251y ago

    Veracode's October 2025 report finds GPT-5 models lead secure coding benchmark

    Veracode's October 2025 GenAI Code Security Report found that OpenAI's GPT-5 and GPT-5-mini reasoning models achieved secure coding rates of 70% and 72%, the highest results among models tested. The report also noted weaker or more moderate performance from non-reasoning models and competing models from vendors including Anthropic and xAI.

  4. Jul 1, 20251y ago

    Veracode begins benchmarking LLM secure coding performance

    Since July 2025, Veracode tested large language models on their ability to avoid introducing SQL injection, weak encryption, cross-site scripting, and log injection flaws across Java, Python, C#, and JavaScript.

  5. May 8, 20234y ago

    Formal study finds widespread vulnerabilities in AI-generated code

    A study titled "Broken by Default" evaluated 3,500 code artifacts from seven widely used LLMs across 500 security-critical prompts and found 55.8% contained at least one vulnerability. Researchers used the COBALT pipeline and Z3 SMT solver to formally prove 1,055 findings, concluding that even the best-performing model still produced insecure code at high rates.

    Broken by Default: A Formal Verification Study of Security Vulnerabilities in AI-Generated Code - Infosec.Pub
Linked entities

Related entities

Vulnerabilities, threat actors, malware, products, organizations, and breaches Mallory has linked to this story.

Open in app →

Organizations

1 linked
Openai
Do you run this?Follow how adversaries are adapting to this technology, and where it touches your stack today.
Track This Risk
Mallory platform

See the full picture, correlated to your attack surface.

This page covers what’s public. Mallory adds the parts that aren’t: which of your assets are affected, which threat actors are using it right now, and what to deploy next.

Exposure mappingMap indicators from this story to your assets and identify affected systems in minutes.
Threat actor evidenceEvery observed campaign, victim, and pivot linked to actors named in this story.
Associated malwareMalware, exploits, and IOCs connected to the activity described here.
Detection signaturesYARA, Sigma, and Snort rules deployed to your SIEM as soon as they’re published.
Scheduled alertsGet matching new stories delivered to your team as they break, not the next morning.
AI threadsAsk questions about this story and take action on the answers.