Media / Press

Media &
Press

This page gathers the public record, citable sources, and contact information for editors, reporters, researchers, and policy readers following the current research programme.

Current research focus

The programme studies what happens on the user side of prolonged AI interaction: how conversational context shapes interpretation, trust, dependence, and judgment over time.

Public records

  • ORCID: 0009-0002-6597-7245
  • SSRN and DOI records for the paper sequence
  • Public methodological documents and repository links
  • Direct contact through rzvn.io

What the current work studies

Current AI safety discourse is heavily system-side. It focuses on alignment, output control, jailbreak prevention, and model behavior. This programme argues that such work is necessary but not sufficient, because prolonged human-AI interaction can also reorganize trust, attachment, and judgment on the user side.

The public record on this site therefore begins with conversational context, moves into user-side contextual hallucination, then into post-interaction assessment and bounded engineering translation. The work is published as observational and methodological research, not as clinical diagnosis or commercial safety certification.

Primary sources

  • CXC-7 for conversational context
  • USCH for the core phenomenon and distinction from model hallucination
  • USCH 14 for the observable phenomena list
  • USCI for public boundary, verifiability, and post-interaction assessment
  • Papers & Notes for the full record

Public Verifiability

Citable, inspectable,
and bounded

The public materials are released for scrutiny, replication attempts, and informed citation. DOI and SSRN records provide versioned citation anchors; methodological documents state their public boundary, controlled-access boundary, and non-clinical scope directly.

Open Papers & Notes → Open Public Notes

Scope

Non-clinical and
methodological

USCH, USCI, and A-CSM are presented as observational and methodological frameworks. They are not clinical instruments, not diagnostic tools, and not high-stakes decision systems for unsupervised deployment.

Current public materials

  • AI Context / CXC-7
  • CXOD-7 and Coh(G)
  • USCH and the 14 phenomena
  • USCI methodology
  • A-CSM technical materials
  • Public notes and briefings

Contact

Email: zon@rzvn.io

Signal: @rzvnzon

ORCID: 0009-0002-6597-7245

Download Press Brief (PDF) Download Headshot

Researcher

ZON RZVN

ZON RZVN, independent researcher based in Taiwan. Non-traditional background in art, music, and tattooing. Entered AI research through self-directed study beginning in late 2024. Since 2025, published three preprints and two technical reports on user-side contextual risk in prolonged human-AI interaction.

Photo available for editorial use with credit: "ZON RZVN"

What this research is NOT

  • Not a clinical or diagnostic methodology
  • Not peer-reviewed (preprints and technical reports)
  • Not externally validated
  • Not a commercial product or service

Why now

A 2026 Stanford-affiliated study (Moore et al., Characterizing Delusional Spirals through Human-LLM Chat Logs, arXiv: 2603.16567) analyzed 391,562 messages and found structured patterns of escalating delusional content in AI user interactions. That study examined what the model produces. This research line examines what patterns manifest on the user side. The two are complementary.

The EU AI Act's phased enforcement reaches key thresholds in 2026. User-facing AI applications in emotional support and companionship categories face scrutiny under high-risk criteria. The user-side risk surface is not addressed by current model-output-centered evaluation frameworks.