Licenses · Legal

AIGENCY V4 licenses, attribution and open-source roadmap.

This page consolidates the legal terms for the AIGENCY V4 Technical Whitepaper, model and services. In case of doubt, this page is authoritative.

Whitepaper Model & Service Component Open-Source Trademark Patents & IP Citation Contact

01Whitepaper License

CC BY-ND 4.0

The AIGENCY V4 Technical Whitepaper (PDF, page content and figures) is licensed under the Creative Commons Attribution-NoDerivatives 4.0 International license. Publisher: eCloud Yazılım Teknolojileri.

You Are Free To

Share the document as-is (PDF, link, quote).
Use in any medium including commercial (report, deck, training).
Cite and reference academically.
Embed page figures unchanged.

You May Not

Publish derivative, adaptation or remix (NoDerivatives clause).
Cut content and redistribute as a new compiled document.
Use without attribution or hide the authorship.
Use AIGENCY/eCloud trademarks to imply endorsement.

Required attribution example

Attribution — plain text

"AIGENCY V4 Technical Whitepaper" by eCloud Yazılım Teknolojileri, licensed under CC BY-ND 4.0. https://aigency.dev/whitepaper-v4

Full license text: creativecommons.org/licenses/by-nd/4.0/legalcode · Human-readable summary: creativecommons.org/licenses/by-nd/4.0

AI Training Data

Using this document as training data for another AI model produces derivative AI weights and requires explicit permission under the 'NoDerivatives' clause of CC BY-ND 4.0. For commercial AI training, please contact [email protected].

02Model & Service License

AIGENCY License Agreement v1.0 (proprietary)

AIGENCY V4 model weights, the trained weight family (including assistant slugs), HBM memory layer and API services are offered under a proprietary license by eCloud Yazılım Teknolojileri. The agreement on the Terms of Service page applies for connecting to the service.

Personal use

Free tier

Commercial use

Per plan

Reselling

Written approval

·API calls consume the user's plan quota; copyright on outputs belongs to the user.
·Model weights, parameter dumps or output samples may not be used as retraining data.
·Reverse engineering, decompilation or parameter extraction via prompt-injection is prohibited.
·Data sovereignty: operates under KVKK §5/§12, ISO/IEC 27001 and EU AI Act high-risk class.

03Component Open-Source Roadmap

Modular and auditable.

The AIGENCY V4 architecture is open-sourced step by step. Each component carries the license most appropriate for its use. Read-only access for academic audit is available from Q3 2026.

Component	License	Release	Note
Training pipeline	`Apache-2.0`	2026/Q3	Full pipeline excluding PII redaction
HBM/CCW reference	`AGPL-3.0`	2026/Q4	64K-token limited reference impl
Vision encoder reference	`AGPL-3.0`	2027/Q1	Excluding Turkish fine-tune data
Cross-modal projection	`AGPL-3.0`	2027/Q1	Full source
Router-Bus & Adapter API	`MPL-2.0`	2026/Q4	Module studio
Benchmark infrastructure	`MIT`	2026/Q3	Reproducibility — runner, scorer, raw logs
Model card signing tools	`MIT`	Preserved	Go + Bash

License rationale

Apache-2.0: patent stability for enterprise adoption. AGPL-3.0: copyleft including network use. MPL-2.0: file-level weak copyleft. MIT: maximum reach.

Excluded

Training pipeline PII redaction rules, vision encoder Turkish fine-tune data, RLHF evaluator pool — closed for user data protection.

04Trademark & Logo

AIGENCY®, eCloud®

Permitted

· Use the AIGENCY name in factual references.
· Display the official logo unmodified in press and news.
· Use descriptive phrases like "integrates with AIGENCY" or "powered by AIGENCY".

Not Permitted

· Use logo or name in your own product ("AIGENCY Pro", "My AIGENCY").
· Imply official endorsement or partnership.
· Modify the logo's colors, proportions or typography.
· Register confusing domains (aigency-x.com etc.).

05Patents & Intellectual Property

Proprietary algorithms protected.

Several AIGENCY V4 architecture components are proprietary algorithms with eCloud Yazılım Teknolojileri patent applications or registrations. Their description in the whitepaper is for academic transparency; commercial reimplementation requires an additional license.

Algorithm	Description
`ZeNO-3`	Zero-Redundancy Node-Optimised distributed training algorithm
`L-MoE`	Localised Mixture-of-Experts routing
`HBM + TG-Decay`	Hierarchical Memory Architecture with time-guided decay
`CCW`	Contextual Core-Wrapping context packaging
`Adaptif LoRA+`	Contextual density metric driven adaptive-rank LoRA

For patent and trademark inquiries: [email protected]

06Citation

How to cite this work

Use the following formats when citing the AIGENCY V4 Whitepaper in academic or commercial documents. The same citation applies to your Hugging Face, GitHub or arXiv mirrors.

BibTeX

@techreport{aigency2026v4,
  title       = {AIGENCY V4: Sovereign, Fully Independent and Multimodal
                 128-Billion-Parameter AI Architecture},
  author      = {{eCloud Yazılım Teknolojileri}},
  institution = {eCloud Yazılım Teknolojileri},
  type        = {Technical Whitepaper},
  number      = {v1.0},
  year        = {2026},
  month       = {April},
  url         = {https://aigency.dev/whitepaper-v4},
  note        = {CC BY-ND 4.0}
}

APA · Turkish

eCloud Yazılım Teknolojileri. (2026). AIGENCY V4: Egemen, tam bağımsız ve multimodal 128 milyar parametreli yapay zekâ mimarisi (Teknik Whitepaper v1.0). https://aigency.dev/whitepaper-v4

APA · English

eCloud Yazılım Teknolojileri. (2026). AIGENCY V4: Sovereign, fully independent and multimodal 128-billion-parameter AI architecture (Technical Whitepaper v1.0). https://aigency.dev/whitepaper-v4

07Contact

License inquiries

License & Legal

[email protected]

General & Product

[email protected]

eCloud Yazılım Teknolojileri · Şanlıurfa, Turkey · Accepts service via registered electronic mail (KEP). This page was last updated on 28 April 2026.

This page is informational and does not constitute legal advice. In case of conflict, the relevant license text (CC BY-ND 4.0 legalcode, Apache-2.0, AGPL-3.0, MPL-2.0, MIT) and the Terms of Service agreement prevail.