The cartographer's table for AI governance.
Lattice helps teams compare, contrast, and apply six major ethical AI frameworks — without flattening the meaningful differences between human-rights ethics, classification taxonomies, certification methodology, Indigenous data sovereignty, adversarial security, and risk management.
Lattice does not provide legal advice, issue certifications, or produce OCAP® determinations. Every claim on screen carries a source, version, and last-reviewed date.
- UNESCONormative ethics grounded in human rights, dignity, transparency, fairness, and human oversight.
- OECDClassify AI systems across People & Planet, Economic Context, Data & Input, AI Model, Task & Output.
- IEEE CertifAIEdCertification methodology for ethical implications of autonomous intelligent systems: transparency, accountability, algorithmic bias, privacy.
- OCAP®First Nations principles for data governance, self-determination, and stewardship.
- MITRE ATLASLiving knowledge base of adversary tactics and techniques targeting AI systems.
- NIST AI RMFVoluntary framework to manage AI risks across Govern, Map, Measure, Manage functions.
Honest cartography
Show relationships, not equivalences. Lattice rejects the string ‘equivalent’ outright. Disagreement between frameworks is information, not noise.
Provenance always visible
No claim without a source. Every excerpt links to the canonical document, with version and last-reviewed date on screen.
Stewardship over speed
For OCAP® and adversarial security content, friction is intentional. The tool routes to community engagement and editorial review, not auto-determinations.
Three modes, one content model
Explore the frameworks, compare them across consistent lenses, or assess a real AI system end to end.
Six frameworks, six different jobs.
They're not interchangeable. Lattice positions each along the axes that actually matter: from values to controls, from voluntary to enforceable, from global to community-specific.
Open the comparisonWhat ships in the workspace
P0· available nowBring a system to the table.
Run the intake, see how every framework reads it, and walk out with a defensible draft. Pre-loaded with four sample systems so you can try it without signing up.