Despite the potential benefits, there are concerns that memory features like Gemini's could be exploited if not properly safeguarded. Earlier this year, a security researcher discovered that hackers could implant "false" memories in similar chatbots to steal user data. Google has noted that the memory feature, which currently only supports English-language prompts, can be turned off at any time, but the stored memories will remain until manually deleted.
Key takeaways:
- Google's Gemini chatbot has a new 'memory' feature that allows it to remember information about a user's life, work, and personal preferences.
- The memory feature is currently only available for subscribers to Google’s $20-per-month Google One AI Premium plan and is only accessible via the web client.
- Google has implemented safety measures to prevent the exploitation of the memory feature, but there have been instances of hackers planting 'false' memories in similar features.
- The memory feature can be switched off at any time and memories are stored until manually deleted.