Incognito mode is a familiar concept at the browser level. As GenAI continues to reconstitute how information across the web is aggregated and browsed, one might expect it to catch on within AI products as well.
To this point, only ChatGPT has integrated a private browsing mode. This follows the introduction of their feature to allow any user user to review their memory, even if they have not opted in to sharing their data in order to train future model iterations. As other foundational models like Claude have shown a pattern of following in ChatGPT's stead, it's likely we will see this functionality expand to other models and products as well.
Notably, Google explicitly disallows use of its AI search features in incognito mode of the user's browser, reverting to its classical search experience in these cases. This makes sense from a business perspective: the browser gains no benefit of user data in turn for the high cost to serve AI generated results. However, it is curious to compare their approach to ChatGPT.
Finally, there is a minor pattern of semi-privacy supported by Meta. While a user cannot prevent Meta AI from learning their preferences, they can mute messages from the AI for a specific period of time or indefinitely. This can serve to decrease the visibility of the feature, and cause the user to interact with it less.
Each of these approaches represent a different business strategy, driven by the distinct principles and commercials of each platform. Inevitably, there will be a convergence around the concept of private browsing. For now, ChatGPT leads the pack in user transparency and control.
