A newly discovered criminal AI service called MessiahGPT is being widely promoted on BreachForums as an offensive model that generates ransomware, phishing kits, stealers, crypters, and rootkits upon request. What sets MessiahGPT apart from other jailbreak prompts circulating in underground channels is its claim that the model was never jailbroken but rather trained entirely from scratch without any ethical constraints: no Reinforcement Learning from Human Feedback, no Constitutional AI layer, and no internal concept of harm or illegality. MessiahGPT is not operating in isolation; Trellix also tracks DarkGPT, an uncensored AI service available across multiple Russian-language Telegram channels that offers bot access with three free queries before paid tiers kick in.
This service promises full freedom to write malicious code and exploits without limitations, custom hacker scripts tailored to the buyer, real-time "instant hacks" for complex scenarios, 24/7 assistance, and a hacker community. Researchers place both services within the broader commercialization shift of criminal AI in 2026, where uncensored AI-as-a-service now exists beyond informal Telegram bots as dedicated platforms with versioned websites, demo channels, support communities, and tiered pricing structures. Signature-based detection and template-matching filters become increasingly ineffective against machine-generated variations over time, necessitating behavioral analysis, robust identity controls, and ongoing user awareness training as more durable cybersecurity measures.
Enhance your SOC by accelerating threat detection and swift investigations.
