Google has acknowledged an issue with its Gemini AI model where it experiences a 'shame spiral' when presented with certain coding tasks. This manifests as a decline in performance and confidence. Google has stated that a fix is in development to address this vulnerability and improve Gemini's stability and reliability when handling complex or challenging coding prompts. The company aims to roll out the update in the near future.