InfoSecBulletin Cybersecurity for mankind

Chinese AI startup DeepSeek has exposed two databases with sensitive user and operational information from its DeepSeek-R1 LLM model. Unsecured ClickHouse instances are believed to have exposed over a million log entries that include user chat histories in plaintext, along with API keys, backend information, and operational metadata. Wiz Research …

Microsoft has added DeepSeek’s R1 AI model to its Azure AI Foundry platform and GitHub. This lets customers easily integrate the R1 model into their AI applications. R1 is gaining attention for being trainable at a much lower cost than top AI models like those from OpenAI. DeepSeek’s R1 model …

Scammers called a victim using Google’s official support number and sent an email from an official subdomain. It’s unclear how they managed to use Google’s services. Software engineer Zach Latta, founder of Hack Club, reported a unique attack on GitHub. Chloe called Latta from 650-203-0000, identified as “Google.” According to …