Artificial intelligence is known for its accuracy issues, so it’s perhaps unsurprising that using it as a coding assistant also generates more security problems. Research from a security firm, Apiiro, shows that developers using AI produce ten times more security issues than those who don’t. Examining code from numerous developers and repositories, Apiiro discovered that while AI-assisted developers generate code faster, they also open up security vulnerabilities at a higher rate.
“It multiplies every type of vulnerability,” stated Apiiro’s product manager, Itay Nussbaum. Paradoxically, some benefits of AI coding contribute to these issues. Syntax errors decreased by 76%, and logic bugs, or faulty code, went down by 60%, but privilege escalation incidents soared by 322%, and architectural design problems increased by 153%.
“In essence, AI is fixing mistakes but creating major security risks,” Nussbaum noted. AI coding’s security drawbacks aren’t a surprise; a study from the University of San Francisco and others flagged similar concerns, finding AI “improvements” degrade overall security.
The data highlights how significant the problem is. With companies like Coinbase, Shopify, and Duolingo requiring AI use, these vulnerabilities not only multiply but also increase workload for those fixing them. AI’s workplace “transformation” primarily results in more tasks for human workers to manage.
While AI’s integration into coding and other fields like writing, audio, and video is advancing, this indicates the issue will worsen before improving.
