ChatGPT now operates without the need for an account — but there’s a twist.

OpenAI is extending access to its primary conversational AI tool to all users, even those who haven’t registered for an account. However, the experience may differ slightly, and all interactions will still contribute to the AI’s training data unless users choose to opt out.

Starting today in select markets and gradually expanding globally, users visiting chat.openai.com will no longer need to log in.

While users can engage in conversations freely, they should note that certain features available to account holders, such as saving or sharing chats and using custom instructions, will not be accessible.

Nevertheless, users retain the option to decline their conversations being used for training purposes, although this may seem contradictory to OpenAI’s initiative. OpenAI provides a helpful gif demonstration for reference.

Even more significant is that this additional free version of ChatGPT will feature “slightly more restrictive content policies.” What exactly does that entail? When asked, a spokesperson provided a verbose yet largely ambiguous response.

The signed-out experience will incorporate the existing safety measures inherent in the model, which already include precautions to prevent the generation of harmful content. Furthermore, we are introducing additional safeguards tailored to address other types of content that may not be suitable for a signed-out experience.

Our approach involves careful consideration of potential misuse scenarios for a logged-out service, drawing on our comprehensive understanding of GPT-3.5 capabilities and risk assessments conducted by our team.

So… it seems there’s no clear indication of what exactly these more restrictive policies entail. “We recognize that additional iteration may be needed and welcome feedback,” the spokesperson mentioned. And they’ll certainly receive plenty of feedback!

Just imagine: a platform whose usage could potentially lead to financial losses for a billionaire. After all, inference remains costly, and even the streamlined, resource-efficient GPT-3.5 model consumes power and server resources. People are likely to push its limits to the fullest extent.

Regarding this concern, they provided a verbose response that lacked any concrete information. It appears they may have as little insight into the potential misuse of the platform as anyone else and will likely need to react rather than proactively anticipate issues.

As for the rollout of ultra-free ChatGPT, it’s unclear which areas or groups will gain access first.