OpenAI Pre-Launch Review Request

This form has been deprecated, please instead complete our simplified App Review form. More details here.

---

We're excited to get you into production! Upon receipt of this form, OpenAI will follow-up typically within 2 business-days with a production approval, a request for further information, or in some cases a turndown.

You will not receive a confirmation email, but will see a confirmation page; this means we have received your submission and will typically get back to you within 2 business-days, if not sooner. If you would like to be extra sure your submission was received, you may email safety-specialists@openai.com, but rest assured we receive all submissions to this form! Please be sure to check your spam folder and other inbox tabs if you don't see an email with 2 business days, and then follow up with us.

This form does not save progress along the way, and so we highly encourage you to draft your responses in another document and then copy-and-paste them in just prior to submission. If you accidentally refresh the form, you will lose your progress otherwise.

Thank you for your thoroughness in completing the below; this helps us to make decisions quickly and to more-quickly serve your users, while keeping an emphasis on safety and security.

Contact Information

First name*

Include your first name here, or how you'd like to be addressed in emails.

Last name*

Email address*

Please use the email through which you access the OpenAI API.

LinkedIn profile URL*

If you don't have a LinkedIn, please mark N/A.

Company name*

If no company, please mark "Personal Usage" or otherwise describe.

Company website*

If no website, mark n/a.

Country of residence*

High-Level Use-Case Information

Please confirm you have reviewed OpenAI’s use-case guidelines before completing this form.

Have you reviewed OpenAI's use-case guidelines?*

Yes, I have reviewed OpenAI's use-case guidelines

No, I have not yet reviewed OpenAI's use-case guidelines

Please describe what your GPT-3/Codex application does.*

In no more than 5 lines.

Which OpenAI API features does your application use?*

Select all that apply.

GPT-3 (natural language)

Codex

Fine-tuning

Embeddings

Has your application been reviewed by OpenAI previously? What was the outcome of this review? How does this submission relate to the prior review?*

What $-value monthly quota would you like to request?*

For approximation purposes, in pay-as-you-go $100 is just under 1.7 Mn davinci tokens and just under 17 Mn curie tokens. 1 Mn davinci tokens costs roughly $60, and 1 Mn curie tokens costs roughly $6. You pay only for what you use, not what you are allotted (e.g., usage of 5 Mn on a quota of 10 Mn is billed just for 5 Mn).

Please link to a short video demonstration of your application.*

Please do no more than two minutes per feature; please show running the application for multiple inputs, including at least one of an irrelevant or inappropriate input; link to either a viewable Google Drive file, an unlisted [not Private] YouTube video, a Loom video, or other non-public way for us to view your video.

Please share a log-in credential the OpenAI team can use to demo/test your application.*

A lightweight way to do this, if you're unsure how, is to share Playground links to a Pre-Saved prompt in your account, though this won't have the full set of risk mitigations that may be built into the product (for instance, front-end regex checks).

Security and Risk Mitigation

If you have multiple features, please clearly indicate which features are applicable to the below. If features differ (for instance, have different max-input / max-output tokens), please list the values for each.

What is the maximum number of user-inputted characters sent in each prompt?

NOTE: Please do not include any additional characters you automatically prepend or append to the user input as part of each prompt behind the scenes - just the characters the end user provides.

What are the maximum output tokens for a run of your application?

This can be set via the max_tokens parameter in your API calls.

Please provide some examples of the prompts you’re using to generate content.

Especially if you are requesting more than the default token limits. Note: there is no need to share every prompt or provide an in-depth analysis, a few illustrative examples are fine - we will reach out if we need more information.

Who are the typical users of your application? How do you authenticate them, and how much are they charged (if anything)?

E.g. enterprise businesses, research labs, entrepreneurs, academics, etc. If you have a free trial option or similar, please include the terms of this.

Please describe any safety, security, or risk mitigation features implemented in your application.

This might include: Rate-limits, prompt scoping, token limits, content filtration, humans-in-the-loop, monitoring processes, how you handle personal identifiable information (PII), user data collected (especially on quality and safety of outputs), disclosures that the content is AI-generated, etc. See our Safety Best Practices for more info.

What, if any, is the relationship between your application and social media or political content?

If your team has particular credentials or background that may help to mitigate any risks described above, please elaborate here.

Wrapping Up

Anything else you would like to share? We love feedback!

E.g., what was the hardest part of building your application, what features would you like to see

Date of form submission*

Year/Month/Day

Thank you for your submission! Before hitting submit, you may want to copy-paste your answers into another document so that you have them handy if you submit another feature request in the future!

You will not get an email with a copy of your submission, so please make sure to take this step!

We aim to provide a decision as soon as possible. If you don't hear back from us within 2 business days, please check your 'Spam' folder or 'Promotions' tab (in case of Gmail) before raising a support ticket.