Analysis Reveals 175,000 Exposed Ollama Instances Could Enable LLM Abuse

An analysis has revealed that 175,000 Ollama instances exposed on the internet could allow for the abuse of large language models (LLMs). Among these hosts, 23,000 have been identified as responsible for the majority of the activity observed over a 293-day scanning period. No specific vulnerability (CVE) or additional technical details are mentioned. The exposure of these instances could facilitate malicious uses of LLMs.