Are criminals vibe coding malware? All signs point to yes

Palo Alto Networks Unit 42 warns that attackers are increasingly using AI "vibe coding" to write and augment malware—sometimes embedding LLM API calls directly into malicious code—and that these tools introduce both realistic threats and flawed "security theater" due to hallucinations; Unit 42 recommends the SHIELD framework (Separation of duties, Human-in-the-loop, Input/Output validation, Enforce security helper models, Least agency, Defensive technical controls) to mitigate the risk.