Red Specter Nightfall is an AI-native offensive security framework designed to test and validate defences against AI-specific attack surfaces. It consists of 64 offensive tools (63 publicly available, 1 restricted to law enforcement) accessible via a single CLI installation. The framework covers five primary attack surfaces: - LLM: Targets the model layer with jailbreaks, context poisoning, adversarial suffix attacks, and constitutional AI bypass - AI Agents: Tests agentic execution layers including MCP boundaries, tool-call hijacking, trust chain attacks, and supply chain compromise - Cloud AI: Targets AI infrastructure on AWS SageMaker, Azure ML, and Google Vertex AI, including IAM misconfiguration, container escape, and model artifact poisoning - Mobile: Tests AI agents in Android and iOS applications via API interception, credential extraction, and certificate pinning bypass - Space/NTN: Targets AI agents in satellite infrastructure and 5G Non-Terrestrial Networks The framework includes NEMESIS, an autonomous adversarial AI component that reasons about defences, mutates attack strategies mid-engagement, and avoids repeating the same attack twice. It operates across 13 kill chain phases with 19 pre-built attack chains and over 61,000 tests. Tools can be used standalone via individual pip-installable packages or together through the full NIGHTFALL CLI, which provides engagement management, attack chain sequencing, signed reports, and audit trails. NIGHTFALL is explicitly designed as the offensive counterpart to Red Specter's AI Shield defensive product, with each tool mapped to a specific defensive control.