In the rush to adopt hot new tech, security is often forgotten. AI is no exception

Cisco Talos used Shodan to discover more than 1,100 Ollama servers exposed on the public internet—about 20% of which were actively hosting models—and warned that unsecured LLM deployments enable attacks such as model extraction, jailbreaks, model poisoning/backdoor injection, unauthorized uploads, resource exhaustion, and possible lateral movement; Talos calls for standardized security baselines, automated auditing tools, and better deployment guidance for LLM infrastructure.