Skip to main content
Home/Risks/IBM2025/Harmful code generation

Harmful code generation

Sub-category
Risk Domain

Vulnerabilities that can be exploited in AI systems, software development toolchains, and hardware, resulting in unauthorized access, data and privacy breaches, or system manipulation causing unsafe outputs or behavior.

"Models might generate code that causes harm or unintentionally affects other systems."

Supporting Evidence (1)

1.
"The execution of harmful code might open vulnerabilities in IT systems."

Other risks from IBM2025 (63)