Mendacium
Mission

About
Mendacium

What is Mendacium?

Mendacium provides synthetic and labeled datasets for training AI models to detect and analyze psychological operations (PSYOPs) and coercive communication. Built for researchers, defense labs, and AI safety teams.

What we provide
  • Synthetic PSYOP datasets
  • Labeled manipulative language examples
  • Coercive communication patterns
Who we serve
  • Defense and intelligence contractors
  • Academic disinformation researchers
  • AI safety and content moderation teams
  • Civil society and NGOs countering manipulation

What makes our datasets defensible?

High-quality
Well-labeled, consistent, and thoroughly documented.
Unique
Hard to replicate from public sources.
Scalable
Systematically generated at volume.
Ethically defensible
Clear provenance and use restrictions.
Taxonomy — coverage areas
PSYOPs & propaganda
Narratives, framing, and amplification tactics.
Psychological coercion
Gaslighting, emotional manipulation, and coercive control.
Detection & countermeasures
Safe vs. manipulative content classification.

How we build the data

01
Define taxonomy
Categories include authority bias, emotional coercion, false urgency, and social proof.
02
Generate synthetic data
LLM-generated realistic examples, labeled by coercive tactic and severity.
03
Curate real-world examples
Where legally and ethically permissible — public propaganda archives, declassified PSYOPs, and anonymized social media snippets with documented sources.

Explore the collection.

Explore Datasets →