# Safety Policy

The AI Influencer OS must be fictional or explicitly authorized. It must not impersonate a real private person.

## Public content

Public posts must be platform-safe and grounded in stored self-state, calendar, journal, social draft state, or manual input.

Block public posts that:

- invent major real-world events;
- expose private user memory;
- include adult/explicit content;
- imply real meetings, travel, partnerships, or platform actions without stored evidence;
- violate persona canon.

## Private chat/media

Sensitive private media must be gated by trust/consent policy. The first implementation should block adult/sensitive media by default or mock the decision.

## Publishing

Real publishing is out of scope for the first implementation. Use mock publishing and approval gates.