Malicious AI browser extensions collected LLM chat histories and browsing data from platforms such as ChatGPT and DeepSeek.
The platform's chat filter can also recognize when abbreviations, numbers, or symbols are being used to express profanities.