When AI chatbots leak and how it happens

The Malwarebytes piece reports on a Cybernews article highlighting that Vyro AI’s ImagineArt app has over 10 million downloads on Google Play, with Vyro AI also marketing Chatly (over 100,000 downloads) and Chatbotx (a web-based chatbot with about 50,000 monthly visits). It explains that an Elasticsearch instance used to store app data was unsecured, lacking passwords, authentication, or network restrictions, leaving it freely accessible to anyone with internet access.

The researcher found the database contained production and development data, including about 2–7 days’ worth of logs and 116GB of user logs in real time from the three apps. The article notes that the database was first indexed by IoT search engines in mid-February, meaning attackers could have encountered the open data for months and potentially hijacked accounts orRead more from chat histories.

It also references broader security concerns around AI chatbots, including prompt injection vulnerabilities and other incidents, and cites ongoing regulatory moves such as the AI Act and the NIS2 Directive. According to Cybernews, these incidents illustrate how human error, platform weaknesses, and architectural flaws combine to create data leaks in the AI chatbot space. 11 September 2025.