What is RiskGPT?
- RiskGPT is a new capability in Red Marker that suggests changes or alternative phrasing for specific high risk sentences with risks removed/reduced.
- This capability is offered on an opt-in basis for clients who wish to benefit from it.
Why do I need it?
- Feedback from legal and compliance on marketing content and assets is often inconsistent, subjective or unactionable meaning responding to flagged legal risks in marketing content and assets can be repetitive and time-consuming, and with no guarantee that the updated amends will be accepted.
- Red Marker and RiskGPT changes that.
- 1) Red Marker scans the document and identifies risks – providing marketers with instant feedback on what’s risky in their content and how to fix it.
- 2) RiskGPT functionality provides suggestions for compliant copy to replace the risky content – offering marketers a fast, effective way to amend content in line with compliance and legal requirements.
- 3) Marketers can then submit their asset or copy to compliance or legal, knowing risk is removed/reduced.
I’ve heard of ChatGPT – is it related?
- Red Marker’s RiskGPT is built using OpenAI’s GPT-3.5 family of large language models, via the OpenAI API.
- We use models to classify and understand language and generate suggestions to reduce the risk in your content.
- GPT is an Artificial Intelligence technology that means Generative Pre-Trained Transformer. The generative part relates to its ability to generate text, the pre-trained transformer part relates to the model’s architecture and training data.
How does Red Marker’s RiskGPT work?
- Red Marker uses proprietary models to identify risky content (typically words, phrases and sentences) in marketing documents.
- Red Marker sends the risky content and special instructions to the OpenAI API.
- The OpenAI API follows the instructions and returns the content with risk removed/reduced.
- The suggested changes are shown in the Red Marker App or Microsoft Word / PowerPoint Add-ins, where the user can insert it into the document, or request a new suggestion.
Does OpenAI get any of my data?
- When we instruct the OpenAI API to make content less risky, we send the risky sentence to OpenAI.
- We do not send any other part of your document.
- OpenAI does not use Content provided to or received from their API (“API Content”) to develop or improve their Services. However, they may use Content from Services other than our API (e.g. ChatGPT and Dall-E) to help develop and improve their Services.
Can I use RiskGPT if ChatGPT is blocked in my organization?
- Red Marker RiskGPT is not ChatGPT.
- Red Marker utilizes Generative AI using the OpenAI API on an opt-in basis.
How does Red Marker ensure RiskGPT suggestions are compliant?
- Red Marker identifies the risky content and sends special instructions to the OpenAI API to remove/reduce the identified risks in the document.
- All AI natural language generation models are probabilistic so their results do vary. This is why we give the user the option to request another suggested change if they are not happy with the suggestion.