> ## Content Index > Fetch the complete content index at: https://thecoverwire.com/llms.txt > Use this file to discover other available public pages before exploring further. # Criminals sell AI hacking tool that runs on stolen API keys - URL: https://thecoverwire.com/llmjacking-stolen-ai-keys-cyber-sublimits/ - Published: 2026-10-17T12:30:00.000Z - Updated: 2026-11-02T11:48:08.000Z - Description: Bitsight found the platform for sale on a cybercrime forum, with its usage billed to the customers whose keys were stolen. - Author: CoverWire - Tags: Cyber, #desk-risk Bitsight has found a criminal hacking platform that runs on stolen AI API keys, leaving legitimate customers with usage charges and testing how cyber policies define and limit the loss. The security ratings company found MONST3R-AI for sale on a restricted cybercrime forum, Cyber Insurance News reported on 16 October. Buyers paste in keys stolen from OpenAI, Anthropic or Google customers, set a budget and point the tool at a list of targets. "The API bill lands on the victim whose key was stolen," Bitsight wrote in a blog post on 15 October. Security firm Sysdig named the practice LLMjacking in 2024\. It estimated that a victim could incur more than $46,000 a day in charges, rising above $100,000 for top-tier models. "An entire market and ecosystem has developed around access to LLMs," Sysdig wrote. The MONST3R-AI seller claimed 1,072 findings across 199 scanned hosts, but Bitsight could not verify the figures. The company expects the bring-your-own-key model to spread, alongside more marketplaces for stolen keys. Some policies already address similar bills. Coalition's Service Fraud cover pays direct financial losses when criminals run up charges on an insured's "cloud- and Internet-based services", although a sublimit may apply. Others have started to name the risk. The Financial Times reported in April that QBE had introduced LLMjacking sublimits, at around $250,000 on a policy with $5m of cover. QBE said it was "not retreating" from AI risks, and that an AI event leading to a conventional cyber incident stays within the core policy. The FT also reported that Beazley had proposed terms to limit some AI-related losses. Beazley said the amendments were "still being developed" and were not in active policies. London cyber MGA Assured said most cyber policies did not explicitly cover LLMjacking. It knew of no insurer that had denied a cyber claim because AI was involved. For buyers and brokers, the wording questions are whether service fraud cover names AI services, whether usage charges count as a direct loss, and which sublimit applies. A conventional cyber incident triggered by AI may remain in core cover while the usage bill sits under a much smaller limit. Bitsight recorded a 45% rise in AI-related discussion across 16 forums from February to August, while ransomware mentions fell 49%. It called keyword tracking "an imperfect measure". The figures track discussion, rather than attacks or insured losses. Bitsight's first recommendation is to treat an exposed AI provider key with the same severity as an exposed AWS credential. Related coverage: [Brit adds cyber to Acturis Package for brokers at commercial renewal](https://thecoverwire.com/brit-cyber-launches-on-acturis-package/) and [Old Mutual says provider error may have exposed customer phone numbers](https://thecoverwire.com/old-mutual-provider-configuration-error-customer-data/). ## Sources - [Cyber Insurance News: Criminals run AI on stolen keys. Some insurers now cap the bill.](https://cyberinsurancenews.org/stolen-api-keys-llmjacking-bitsight-2026/?ref=thecoverwire.com) - [Financial Times: Insurers move to cap cyber payouts related to AI and LLMjacking](https://www.ft.com/content/12e36e02-7ff9-4a45-9544-872822fe9c97?ref=thecoverwire.com) - [QBE: AI as a risk amplifier](https://www.qbe.com/us/conversations/ai-as-a-risk-amplifier-how-cyber-insurance-is-evolving?ref=thecoverwire.com)