Windows botnet x47.c uses xAI's Grok API to pick its next move

A newly identified Windows botnet called x47.c uses xAI's Grok API to help it stay on infected machines, according to a report published by SecurityWeek on 26 September 2026. The malware relies on the AI service to choose from a set of predefined actions, and it also drains AI API resources on the systems it infects.

A botnet is a network of compromised computers that an operator controls remotely. Persistence is the term for the techniques malware uses to survive reboots and clean-up attempts so that it keeps running on a machine. According to SecurityWeek, x47.c leans on Grok, xAI's large language model service, to make that decision-making step, rather than following a fixed script written into the malware itself.

The important detail is what the AI is and is not doing here. SecurityWeek says Grok is used to select from predefined actions. The choices available to the malware appear to have been set out in advance by whoever wrote it; the model picks among them. That is a narrower role than writing new attack code on the fly, and it places the case in a category security researchers have been tracking more closely: ordinary malware bolting a commercial AI API onto its control logic.

The second behaviour SecurityWeek flags is AI API draining. That means consuming the AI service credits or quota associated with an infected system. For an organisation, the cost of an infection would not only be the compromised endpoint but also whatever API spend the malware runs up.

The root cause, on the information available, is tool misuse rather than a software flaw. There is no indication in the reporting that a vulnerability in Grok or in the Grok API was exploited. The malware simply uses the service the way any other client would, for purposes the service is not meant to serve.

Details remain limited. SecurityWeek's report does not, in the material available here, state how many systems are affected, how the botnet spreads, who is operating it, or when the first infections occurred. xAI has not been reported as having commented, and no fix, block or account action has been described. The Agentic Times has not independently verified the findings.

Organisations that issue AI API keys to staff or to automated systems may want to check where those keys are stored on endpoints and what spending limits are attached to them. Unusual API usage from a workstation is, on the basis of this report, now a signal worth watching alongside more familiar signs of compromise.