ChatGPT but without the privacy risks? Kind of.

It’s not as if any privacy concerns regarding ChatGPT ever really dissuaded anyone from using it – come on now, they never did – but OpenAI seems to think that for some people this may have indeed been the case, hence its latest announcement regarding the famous service: people will soon be able to use ChatGPT without having to log in, allowing for anonymous access to prompts and results for the first time. As one would expect though – since OpenAI is actually anything but a non-profit – there are certain things people should know if they’re interested in using ChatGPT that way.

It’s not available to everybody yet, for one: OpenAI “will be rolling this out gradually, with the aim to make AI accessible to anyone curious about its capabilities” (people can check whether this login-free version of ChatGPT is available in their country by visiting this page). Then it’s the kind of obvious functionality that just can’t be offered to users who haven’t logged in, such as saving one’s prompt history, viewing past prompts at any time or sharing prompt chats with others in a variety of ways. Both of these limitations are only to be expected for a service of this kind, though, especially when it’s used without an active profile.

There are other noteworthy limitations too. ChatGPT users who access the service without logging in, for instance, will have to live with “additional content safeguards”, OpenAI notes, “such as blocking prompts and generations in a wider range of categories”. These are probably introduced so as to prevent anonymous abuse of ChatGPT’s model – AI abuse is practically a sport of sorts, after all, even among logged users. While OpenAI did not offer any other details, chances are that these safeguards are also put in place with minors in mind, which can only be a good thing (but still limiting).

The most important limitation of this login-free ChatGPT, though, has to do with its version number: that would be the older v3.5, which is infamous for its often erroneous, untrustworthy results or the literally fabricated “facts” included in many of its answers on a number of different subjects.

ChatGPT v4.0 and v4.0 Turbo are much better AI models overall, but they are still locked behind the paywall of a ChatGPT Plus subscription. Even Microsoft’s Copilot AI Assistant – automatically installed through the latest Windows 11 update, whether consumers like it or not – is now based on ChatGPT v4.0 Turbo, so OpenAI’s decision to stick with v3.5 for the login-free ChatGPT might make some financial sense but it’s also a shortsighted one. It’s definitely not helping the company’s case if it truly means to impress newcomers to AI chatbots, that’s for sure.

In terms of data collection, OpenAI had this to say regarding the login-free version of ChatGPT:

We may use what you provide to ChatGPT to improve our models for everyone. If you’d like, you can turn this off through your Settings – whether you create an account or not.

In practice this means that the company’s public stance regarding data gathering has not changed: anonymous users of ChatGPT can still opt-out of the default privacy setting of the service (which uses prompts and results as data for AI model training), just like logged-in or paying users can. Whether OpenAI’s claims about not gathering user data, though, are as empty as Microsoft’s regarding Windows user data is anybody’s guess. It’s not like either company has been caught being untruthful about that kind of stuff before, right? Right?