2025 GenAI Code Security Report Assessing The Security of Using LLMS for Coding

As AI-powered coding assistants become standard in software development, a key question arises: are the large language models behind them generating secure code? Developers often rely on these tools without recognizing security implications, leaving critical decisions to the model.

This report evaluates the security performance of 100+ LLMs across languages and vulnerabilities, offering insights for security and development teams. Key findings:

• Only 55% of AI-generated code tasks were secure across all models
• Security performance stayed flat over time, regardless of model size
• Results varied by vulnerability type, with XSS and log injection performing worst

Read the full report for details.