Research

Samantha Maeer is the founder and creator of contAIn™, the methodology layer that directs AI to the outcome you define. Her research documents how AI governance fails and what holds in its place, evidenced from live production use across multiple projects rather than a test environment. Both papers below are published research, indexed under ORCID 0009-0000-5439-3645.

The Feed Loop: How AI-Generated Governance Documents Amplify the Patterns They Were Designed to Suppress

Samantha Maeer, 2026

AI systems copy patterns at higher volume than they follow prohibitions. When governance files are written by AI, they carry stylistic defaults that contaminate the same files containing rules designed to suppress those patterns. The fix is not stronger prohibition. The fix is purging the contamination source.

This paper documents the discovery and formal naming of a previously undescribed AI failure mechanism, identified through live operational use across multiple project case studies.

Published research. On Zenodo, SSRN and ResearchHub.

DOI: https://zenodo.org/records/20474271
ResearchHub: View on ResearchHub
ORCID: https://orcid.org/0009-0000-5439-3645

A Production Field Report on Human AI Governance: Why Prohibitions Fail and Gates Hold

Samantha Maeer, 2026

The companion paper to The Feed Loop, and the field report behind the contAIn™ method. It answers the question every operator is now asking: why does AI keep ignoring its own rules, and what actually stops it? The answer is not a stronger ban but a structural gate and a positive criterion, the methodology layer that directs AI to the outcome you define.

The record reaches back to December 2025. contAIn's governance and gates were built in live production from a December 2025 origin that predates the failure register itself. An informal prior-art search on 29 June 2026 located no existing work documenting a benign, self-authored prohibition that re-seeds and amplifies the very pattern it bans, so the paper claims this as newly documented ground.

The evidence is a dated, failure-only register: 705 classified failures across 227 sessions, logged between 21 February and 18 June 2026, with 112 human catches recorded. Prohibition-based AI governance fails where governance is broad, personality-heavy and ungated. Gate-and-criterion governance holds where work is rules-based, judged on specific criteria, and enforced by structural gates. The constant across both is the human directing the system rather than deferring to it.

Published research. On Zenodo and ResearchHub.

DOI: https://zenodo.org/records/21206719
ResearchHub: View on ResearchHub
ORCID: https://orcid.org/0009-0000-5439-3645

Both are published research. Further work in progress.