IF YOU THINK ANYONE IN THE AI INDUSTRY HAS ANY IDEA WHAT THEY'RE DOING, IT APPEARS THAT DEEPSEEK JUST ACCIDENTALLY LEAKED ITS USERS' CHATS

Summary from the TLDR AI newsletter

Wiz researchers discovered unencrypted sensitive data in DeepSeek's open-source AI model, exposing major security vulnerabilities. Despite difficulties in contacting DeepSeek, the databases were secured within an hour after the issue was reported. Wiz highlighted that such security oversights indicate the service is not mature enough to handle sensitive data.