Code security leader says artificial intelligence mimics human errors but offers solutions too

LAS VEGAS. Artificial intelligence large-language models mimic human developers in how they write software code, including their tendency to leave behind serious security vulnerabilities, according to Veracode CTO and cofounder Chris Wysopal, who argued that AI itself will be essential to remediating this problem.

“We are working on training LLMs to fix code,” Wysopal said, delivering a twist on AI’s often vaunted ability to write code. “Automated fixing is the only way to fix this, this is a big part...