Why Prompts Are the New IOCs You Didn’t See Coming!

This article analyzes Anthropic’s March 2025 report on malicious uses of Claude and translates it into actionable guidance, arguing that prompts are emerging Indicators of Compromise and that LLM-specific TTPs should be tracked. It outlines likely adversarial prompts across four scenarios—influence operations, credential stuffing against IoT cameras, recruitment fraud, and novice malware development—and supplies sample NOVA pattern-matching rules to detect these behaviors, while noting the lack of concrete IOCs in the original report.