OpenAI Launches ChatGPT to Unregistered Users with Stricter Content Policy

Key Points - ChatGPT is now available to unregistered users without the need to sign-in; accessible at chat.openai

Key Points

- ChatGPT is now available to unregistered users without the need to sign-in; accessible at chat.openai.com.

- Unregistered users will not have access to certain premium features like saving or sharing transcripts.

- All conversations will still be used for training data by default unless users opt-out.

- The free version of ChatGPT will implement a "more stringent content policy" to prevent harmful content.

Details

On Monday, April 2nd, OpenAI announced that their conversational AI, ChatGPT, is now open to unregistered users. This marks a significant step for the platform as it becomes accessible without the need for a login.

It's important to note that all conversations will still be used for training data by default unless users choose to opt-out. Users can disable this feature by clicking on the question mark icon in the bottom right corner and accessing the "Settings" page.

The feature has been gradually rolling out globally since Monday, U.S. time.

Content Policy in the Free Version of ChatGPT

The newly introduced free version of ChatGPT will implement a more stringent content policy. However, when asked for specifics about this policy, the spokesperson provided a lengthy and somewhat convoluted response.

In essence, it suggests that ChatGPT will benefit from the built-in safety measures of the model which should prevent it from generating harmful content. Additionally, some extra safeguards have been added to address certain types of content that may not be appropriate for non-logged-in users.

Measures to Tackle Abuse

When asked if OpenAI has developed strategies to prevent potential large-scale abuse and misuse of the model, the company did not respond. However, the potential risks are evident.

Even the improved, more efficient GPT-3.5 model requires substantial amounts of electricity and server power, which could incentivize bad actors to find ways to exploit it for their purposes.

OpenAI's response regarding how it will handle the threat of abuse for the free version of ChatGPT was similarly vague. The company states that it has put considerable thought into abuse detection and prevention, but it did not provide any concrete details.

This suggests that OpenAI may not fully understand or anticipate the potential risks of misuse for this new service. This could mean that they will adopt a more reactive approach rather than having a comprehensive strategy in place.

The opening up of ChatGPT to unregistered users is a significant development in the field of AI. However, it is important to be aware of the limitations of the stricter content policy and take steps to prevent abuse. It will be interesting to see how OpenAI responds to these challenges as the new service gains traction.