CrowdStrike finds DeepSeek misalignment

Tom Uren and Amberleigh Jack talk about new research that shows the Chinese-made DeepSeek-R1 AI model produces insecure code when prompts include topics that the Chinese Communist Party dislikes. It’s interesting research, but the CCP doesn’t have a monopoly on imposing AI bias.

They also discuss the complete doxxing of the Iranian cyber espionage group known as APT35 or Charming Kitten.

This episode is also available on Youtube.

Show notes