How we use AI: the principles

7 commitments that guide how Mani generates, stores, and delivers AI content.

1

Transparency

Every AI-generated output is labeled where required by platform policy. We do not disguise AI content as human-created when disclosure is expected.

2

Bias mitigation

We review generation outputs across demographics to identify and correct systematic biases in imagery, copy tone, and audience assumptions.

3

Brand safety

Generations stay within your locked Brand DNA. The AI has creative freedom on angle and composition, but zero freedom to deviate from your brand identity.

4

Data privacy

We do not train models on your content. Your Brand DNA, product photos, and generated ads are never used to improve AI models for other customers.

5

Human oversight

Every output passes through the approval queue before publishing. No ad goes live without a human reviewing and approving it. AI generates, humans decide.

6

IP respect

Mani only generates using assets you provide or that are licensed for commercial use. We do not scrape, copy, or remix other brands without authorization.

7

Continuous improvement

We publish quarterly reviews of model behavior, flag rates, and corrections. When we find a pattern of problematic outputs, we fix the system and disclose the change.

What we're working on

Automated bias detection across generated imagery (skin tone, gender representation, cultural sensitivity).

Platform-specific disclosure compliance (auto-labeling AI content where Meta, TikTok, or LinkedIn require it).

Quarterly responsible AI report (first edition targeting Q4 2026).

Related: Content Safety · Security · Privacy · Trust Center