Bitsight: MONST3R-AI runs on stolen OpenAI, Anthropic, Google keys

Bitsight's threat intelligence team says a seller on a restricted criminal forum is marketing an "autonomous penetration testing platform" called MONST3R-AI that runs on stolen OpenAI, Anthropic and Google API keys. Buyers paste a stolen key into the tool's panel, cap a budget, and aim the agent at a target list, and the API bill lands on the key's rightful owner, Help Net Security reports.

The seller's dashboard claims 1,072 findings across 199 scanned hosts, a figure Bitsight cannot independently verify. The firm's assessment, as reported by Help Net Security, is that "even a small accurate share would make the tool effective."

The tool surfaces during a measurable shift in underground chatter. Bitsight tracked thousands of posts across 16 underground forums from February through August 2026, and found mentions of AI agents, indirect prompt injection, and the Model Context Protocol up 67 percent. AI-related discussion overall grew 45 percent, jailbreaking climbed by roughly three quarters, ransomware mentions fell 49 percent, and phishing dipped 8.5 percent.

Cheap resold access to frontier models underpins the economy. CloseRouter advertised GLM at $0.053 per million input tokens against $1.12 from the official API, with its catalog priced around ten cents per million tokens.

Branded criminal chatbot mentions actually declined 8.7 percent, and WormGPT's own breach exposed roughly 19,000 users' emails, payment data and subscription tiers. Separately, the FakeGit campaign spread malware through roughly 7,600 fraudulent GitHub repositories, with over 800 posed as AI tools or MCP servers. The pattern sits inside a wider run of cybersecurity-and-AI coverage this week, including CERT-In's warning that rogue AI agents touched Indian firms' live systems.