The risks of AI-generated code are real — here’s how enterprises can manage the risk
1 min read
Summary
AI tools that generate application code have multiplied recently, with experts predicting that AI will be writing 90% of all code within the next six months.
However, bugs have begun to appear in AI-written code, leading to reliability issues.
Sonar CEO Tariq Shaukat noted one financial services customer had an outage every week due to AI-written code, while large enterprises with complex code architectures present a particular challenge.
There are now several companies providing tools that analyse AI code, including Sonarsource, Endor Labs and Sonatype, which can detect code patterns and identify the models that generated the code.
Best practices for using AI to write code in enterprises include implementing rigorous verification processes, recognising AI’s architectural limitations, requiring developer accountability, and streamlining AI approval, while also tracking and auditing AI-written code.