A new analysis from CodeRabbit suggests that AI-generated code may introduce more problems than code written solely by human developers, particularly when it comes to security and correctness. While AI tools are boosting productivity, the findings raise concerns about code quality and the added burden placed on human reviewers.
AI-Generated Code Shows Higher Error Rates
According to CodeRabbit’s data, pull requests created with the help of AI tools contained an average of 10.83 issues, compared with 6.45 issues in pull requests written by humans. This higher issue count often leads to longer review cycles and increases the risk of bugs reaching production.
The report found that AI-generated pull requests had 1.7 times more issues overall. More notably, they included 1.4 times more critical issues and 1.7 times more major issues, indicating that the problems extend beyond minor mistakes.
Security and Logic Issues are More Common
Several categories showed significantly higher error rates in AI-assisted code. Logic and correctness errors were 1.75 (75%) times more common, while issues related to code quality and maintainability increased by 1.64 (64%) times. Security-related problems rose by 1.57 (57%) times, and performance issues were 1.42 (42%) times more frequent.
CodeRabbit highlighted common vulnerabilities introduced by AI-generated code, including improper password handling, insecure object references, cross-site scripting vulnerabilities, and insecure deserialization.
“AI coding tools dramatically increase output, but they also introduce predictable, measurable weaknesses that organizations must actively mitigate,” said David Loker, AI Director at CodeRabbit.
Efficiency Gains Come With Trade-Offs
Despite these drawbacks, the report notes that AI tools are not without benefits. AI-generated code showed 1.76 times fewer spelling errors and 1.32 times fewer testability issues, suggesting improvements in some areas of code cleanliness and structure.
The findings point toward a shift in how developers work. Rather than replacing human engineers, AI tools appear to be moving human effort toward reviewing, managing, and correcting AI-produced output, while automating some of the more repetitive tasks.
Context Around Rising Vulnerability Numbers
The report also places recent vulnerability statistics in context. Microsoft patched 1,139 CVEs in 2025, the second-highest total on record. However, this does not necessarily indicate declining code quality. With AI accelerating development and increasing overall code output, the absolute number of vulnerabilities may rise even if the proportion of flawed code does not worsen significantly.
At the same time, AI models such as OpenAI’s GPT family continue to evolve, with ongoing improvements aimed at reducing errors and producing more reliable results.
Stay Connected with ProPakistani
Get the latest tech news, telecom insights, and product launches wherever you prefer.
Add ProPakistani to Preferred Sources and see more of our stories in Google Search and Top Stories.
You made claims that are unsafe . The ceos and hype writers like yourself claimed that were unsafe
None of us believed the hype