A publication dated October 5, 2025 reports that prompts found with GEPA detected malicious behavior in AI-generated code, blocking 90% of malicious submissions with 1% of the audit budget.
On October 5, 2025, a publication reported that GEPA found prompts for detecting malicious behavior in AI-generated code. According to the account, those prompts blocked 90% of malicious submissions using 1% of the audit budget.
The publication describes a prompt-optimization use case for evaluating AI-generated code. Consult the original material to verify the method, context, and details behind the figures; the available summary does not explain how the test was conducted. If you use an AI tool to study or apply the method, avoid submitting code, credentials, or internal data without appropriate authorization and protection.