Help Net Security reported on 16 October 2026 that Bitsight, a security firm, found a seller on a restricted criminal forum offering MONST3R-AI. The seller pitches it as an AI agent that tests targets for weaknesses without human steering. A buyer pastes in stolen OpenAI, Anthropic or Google keys, sets a budget cap and lists targets. The seller hosts no model, so the key owner pays the charges.
The report says the setup sidesteps the provider's safety training, but it does not explain how. The seller claims 1,072 findings across 199 scanned hosts. That figure comes from the seller's dashboard, and Bitsight cannot verify it. The report gives no count of stolen keys. Bitsight tracked 16 underground forums from February to August 2026 and saw posts naming MCP, indirect prompt injection or AI agents rise 67 percent.
A leaked key lets someone else run attacks on third parties at the owner's expense. The report advises scanning for exposed AI provider keys and treating a leak like an exposed AWS credential. It names a second exposure: more than 800 of roughly 7,600 fake GitHub repositories spreading SmartLoader malware posed as AI skills or MCP servers. It advises checking where every plugin comes from.