Brazen.AI builds an AI-native marketing platform. This page explains how AI is used in the product, what data flows where, and the commitments we make about how customer information is handled. It complements our Privacy Policy and our Terms of Service.
How we use AI
AI agents within the platform perform tasks that historically required marketing teams: analyzing performance data, drafting campaign strategies, generating creative assets, building landing pages, optimizing bids, and surfacing decisions for human review. AI is integral to the Service; it is not an add-on. Customers configure scope, guardrails, and approval requirements at the agent level.
Models we use
We rely on a combination of third-party foundation models, smaller purpose-built models we fine-tune internally, and deterministic rule engines. Current third-party providers include:
- OpenAI — large-language and multimodal models for drafting, summarization, and tool use.
- Anthropic — large-language models for reasoning and long-context tasks.
- Google — Gemini family models and embedding services for retrieval-augmented workflows.
Provider selection per task is made based on quality, latency, regional availability, and contractual data-handling commitments.
Customer data and AI training
We do not use customer prompts, customer audience data, or end-user personal information to train foundation models. Where we operate third-party models, we configure provider zero-retention agreements wherever they are available — at the time of writing this includes OpenAI zero-retention and Anthropic zero-retention . Where we fine-tune internal models, we do so only on aggregated and de-identified data, and customers can opt out from any model improvement use by contacting privacy@brazen.ai.
Generated content disclaimers
AI outputs may contain inaccuracies, omissions, or claims that require independent verification. Customers are responsible for reviewing AI-generated content before publication, particularly in regulated verticals (health, finance, legal, employment, housing, and credit). Brazen.AI's Decision Inbox surfaces consequential actions for explicit human approval; customers should not bypass approval workflows for content that requires regulatory or compliance review. Where required by law or by the destination ad platform, customers must label AI-generated or AI-altered creative accordingly.
Human oversight & override
Every consequential automated action — for example, publishing a new ad, changing a budget, or launching a landing page — can be gated for human approval through the Decision Inbox. Customers may also override or roll back any AI-driven action. End-users who believe they have been subject to a decision with legal or similarly significant effects can request human review by emailing privacy@brazen.ai; we will respond within the timelines set out in our Privacy Policy.
EU AI Act considerations
Under the EU AI Act, the Brazen.AI platform is preliminarily classified as a general-purpose / limited-risk system . We comply with the Act's transparency obligations by: informing users when they are interacting with an AI system, by labeling synthetic content where required, by maintaining technical documentation appropriate to our risk class, and by providing the information needed for downstream deployers to comply with their obligations.
Bias mitigation
We test models in our pipeline against representative prompts and flag outputs that appear discriminatory or unsafe. We restrict certain categories of advertising — including discriminatory targeting in housing, employment, and credit — at the platform level (see Acceptable Use Policy). We acknowledge that no current bias-mitigation approach is perfect, and we welcome feedback at legal@brazen.ai.
Contact
AI policy and trust questions: legal@brazen.ai
Privacy / data subject requests: privacy@brazen.ai