AI training data
In placeWe do not train or fine-tune any AI model on your data, and we do not license it to anyone for training.
We do not use customer personal data, customer content or output to train or fine-tune any AI model, and we do not license it to any third party for training.
Source: DPA, Section 3.4
AI bias
PolicyWritten bias threat modeling and bias prevention guidelines exist. They have not been independently audited.
Written documents describe how we model bias risks in AI features and the guidelines we follow to reduce them.
This is a written internal policy. It has not been independently audited.
Source: Bias Threat Modeling (on request)
AI security
PolicyA written red team testing policy covers adversarial testing of our AI features.
A written policy describes red team testing of our AI features: its scope, threat categories, method and how findings are fixed.
This is a written internal policy. It has not been independently audited.
Source: Red Team Testing Policy (on request)
AI monitoring
In placeIn Consulting, every AI action that reaches a client is logged with the model, the run id and who approved it.
In Consulting, every AI action that reaches a client is logged with the model, the run id, who approved it and when.
The chat asks for confirmation before anything that changes money or scope.
Source: Consulting security
AI manifest
In placeIn Agency, each deliverable carries an AI manifest that names the models that made it.
Each Agency deliverable carries an AI manifest that names the models that made it. Every model it can name belongs to a listed model provider.
Source: Agency security
Model providers
In placeThe providers of the AI models we run are listed, with where they process data.
Model providers process requests under their own API terms. We select API offerings and settings intended for business use.
In One API you choose the model. In Agency and Consulting, we select the models that do the work.
Source: Subprocessors, Section 4
Human approval of AI output
In placeIn Agency a person checks each deliverable and you approve it. In Consulting you sign off each delivery.
In Agency, a person at PlungeAI checks deliverables, and you approve each one before it is used. In AI Consulting, you sign off each delivery.
These reviews reduce the risk of error but do not remove it. You remain responsible for reviewing output before you rely on it.
Source: Terms of Service, Section 8.3
No automated decisions
In placeWe make no decisions about people based solely on automated processing.
We do not make decisions based solely on automated processing that have legal or similarly significant effects. On the waiting list, an AI model suggests a score and a person makes every decision.
Source: Privacy Policy, Section 5.4