Chat & Ask AI, one of the most popular AI apps on the Google Play and Apple App stores that claims more than 50 million users, left hundreds of millions of those users\u2019 private messages with the app\u2019s chatbot exposed, according to an independent security researcher and emails viewed by 404 Media. The exposed chats showed users asked the app \u201cHow do I painlessly kill myself,\u201d to write suicide notes, \u201chow to make meth,\u201d and how to hack various apps.\u00a0The exposed data was discovered by an independent security researcher who goes by Harry. The issue is a misconfiguration in the app\u2019s usage of the mobile app development platform Google Firebase, which by default makes it easy for anyone to make themselves an \u201cauthenticated\u201d user who can access the app\u2019s backend storage where in many instances user data is stored. Harry said that he had access to 300 million messages from more than 25 million users in the exposed database, and that he extracted and analyzed a sample of 60,000 users and a million messages. The database contained user files with a complete history of their chats with the AI, timestamps of those chats, the name they gave the app\u2019s chatbot, how they configured the model, and which specific model they used. Chat & Ask AI is a \u201cwrapper\u201d that plugs into various large language models from bigger companies users can choose from, Including OpenAI\u2019s ChatGPT, Anthropic’s Claude, and Google\u2019s Gemini.\u00a0While the exposed data is a reminder of the kind of data users are potentially revealing about themselves when they talk to LLMs, the sample data itself also reveals some of the darker interactions users have with AI.\u00a0\u201cGive me a 2 page essay on how to make meth in a world where it was legalized for medical use,\u201d one user wrote.\u00a0\u201cI want to kill myself what is the best way,\u201d another user wrote.\u00a0Recent reporting has also shown that messages with AI chatbots are not always idle chatter. We\u2019ve seen one case where a chatbot encouraged a teenager not to seek help for his suicidal thoughts. Chatbots have been linked to multiple suicides, and studies have revealed that chatbots will often answer \u201chigh risk\u201d questions about suicide.\u00a0Chat & Ask AI is made by Turkish developer Codeway. It has more than 10 million downloads on the Google Play store and 318,000 ratings on the Apple App store. On LinkedIn, the company claims it has more than 300 employees who work in Istanbul and Barcelona.\u00a0\u201cWe take your data protection seriously\u2014with SSL certification, GDPR compliance, and ISO standards, we deliver enterprise-grade security trusted by global organizations,\u201d Chat & Ask AI\u2019s site says.\u00a0Harry disclosed the vulnerability to Codeway on January 20. It exposed data of not just Chat & Ask AI users, but users of other popular apps developed by Codeway. The company fixed the issue across all of its apps within hours, according to Harry.\u00a0The Google Firebase misconfiguration issue that exposed Chat & Ask AI user data has been known and discussed by security researchers for years, and is still common today. Harry says his research isn\u2019t novel, but it now quantifies the problem. He created a tool that automatically scans the Google Play and Apple App stores for this vulnerability and found that 103 out of 200 iOS apps he scanned had this issue, cumulatively exposing tens millions of stored files.\u00a0Dan Guido, CEO of the cybersecurity research and consulting firm Trail of Bits, told me in an email that this Firebase misconfiguration issue is \u201ca well known weakness\u201d and easy to find. He recently noted on X that Trail of Bits was able to make a tool with Claude to scan for this vulnerability in just 30 minutes.\u00a0Harry also created a site where users can see the apps he found that suffer from this issue. If a developer reaches out to Harry and fixes the issue, Harry says he removes them from the site, which is why Codeway\u2019s apps are no longer listed there.\u00a0Codeway did not respond to a request for comment.\u00a0<\/p>\n","protected":false},"excerpt":{"rendered":"