Trust, Privacy & Safety
Built for transparent and privacy-first moderation.
AEGIS is designed to protect users from harmful advertising while minimizing unnecessary data collection, supporting local-first analysis, and making moderation decisions explainable.
Trust Principles
The values behind every decision.
AEGIS is built around a clear set of principles that guide how content is detected, scored, and moderated.
Privacy-first
Collect less data and process locally where possible. Remote services are opt-in and clearly disclosed.
Transparency
Make policies, categories, and decisions understandable to developers, users, and the public.
User control
Let developers, parents, and organizations configure enforcement to match their audience's needs.
Explainability
Provide clear reasons for labels, blurs, and blocks so users can understand and appeal decisions.
Accountability
Design the core engine for public review, security auditing, testing, and continuous improvement.
Proportional enforcement
Use labels for uncertainty, blurs for risk, and blocks only for clearly high-risk content.
Privacy Architecture
Local-first, transparent by default.
AEGIS should avoid sending sensitive browsing data to external services unless a user, developer, or organization explicitly enables cloud features. The system is designed to support local-first safety analysis wherever practical.
What this means in practice
- Local-first does not mean every feature will always run offline.
- Some enterprise or advanced features may require hosted services.
- The product will be transparent about what is processed locally and what is processed remotely.
Explainability
Every decision comes with a reason.
When AEGIS moderates content, it should provide understandable reasons so users and developers can understand, appeal, or configure the outcome.
This ad was blurred because it contains suspicious financial claims.
This promotion was labeled because the destination domain appears risky.
This content was blocked because it matches a child-safety policy.
False Positive Awareness
Safety systems can make mistakes.
AEGIS should be designed with feedback mechanisms, override controls, policy tuning, and review workflows to reduce unnecessary blocking of legitimate content.
- Do not overclaim accuracy or detection rates
- Allow configurable enforcement levels per context
- Prefer labels for uncertain cases
- Reserve blocking for clearly high-risk content
- Support feedback and review paths
Safety Taxonomy
The categories AEGIS is designed to detect.
AEGIS works across a defined set of harmful advertising categories. These categories guide detection, scoring, and policy enforcement.
Phishing and malware
Promotions that attempt to steal credentials or drive users toward unsafe downloads.
Financial scams
Suspicious investment offers, guaranteed returns, and deceptive financial promotions.
Fake health claims
Unverified medical promotions, miracle-cure advertising, and dangerous wellness misinformation.
Counterfeit products
Ads for replica goods, unlicensed items, and unauthorized brand impersonation.
Adult or exploitative content
Age-restricted or exploitative material appearing outside appropriate contexts.
Child-unsafe promotions
Content inappropriate for minors including gambling-style mechanics and unsafe offers.
Deceptive AI-generated ads
Synthetic media and AI-generated creatives designed to mislead or impersonate.
Suspicious sponsored links
Search or display placements linking to high-risk or reputation-flagged domains.
Build on a foundation of trust.
Explore how AEGIS approaches responsible moderation, privacy-first infrastructure, and open-core transparency.