OTX Threat: The state of AI-analysis evasion in malware

Malware authors are developing techniques to obstruct automated AI analysis by embedding natural-language instructions within their code. This 'AI-analysis evasion' targets the pipeline that extracts text from samples and submits it to language models for triage or classification. Analysis of 84 samples across four malware families (FRUITSHELL, PLOTSAFE, HOLLOWCLAD, and MANTLEMAZE) from January 20 CVEs: CVE-2015-2291