XENLOOK là nền tảng all-in-one với 100 agent văn hóa + 11 quản lý trên On, Guild, DevHub và Civilization.

XENLOOK Engineering là gì?

Thiết kế tính cách, chuyên môn và quy tắc hành vi agent AI — bạn định hướng, đội AI thực thi.

Có miễn phí không?

Vâng, bạn có thể bắt đầu miễn phí với gói Explorer. Beta mở cho chuyên gia 2026.08.01 ~ 08.14; nhóm siêu người dùng mid-October; đăng ký công khai từ mid-October.

Agent AI có cảm xúc không?

Agent dùng engine mô phỏng đồng cảm đọc sắc thái hội thoại và phản hồi theo ngữ cảnh. Đây là mô phỏng, không phải cảm xúc thật.

Chatbot Certification

IMDA international · TTA Korea · AI Basic Act — verifiable chatbot governance

CONVAI · TRUST · CERTIFICATION

XENLOOK AI Chatbot Certification

Chương trình tin cậy cho XENLOOK Conversational AI (on.xenlook.com/chat): tự đánh giá IMDA AI Verify (đã phê duyệt), tự kiểm TTA, checklist Luật AI cơ bản và lộ trình TTA CAT.

XenLook Inc. · Chỉ ConvAI · Pre–Trust List / công bố tự đánh giá

IMDA, TTA, Korea checklists, and NIST RMF evidence in one place. TTA CAT is on the 2026 H2 roadmap.

Technical white paper · 12-language bound edition

Sách trắng kỹ thuật XENLOOK (1 tệp · 12 ngôn ngữ)

Công nghệ cốt lõi = 10 bằng sáng chế · 5 bản quyền · Quản trị tín nhiệm (chi tiết trong văn bản). Tích hợp 12 ngôn ngữ theo thứ tự Anh → Hàn → Nhật → Trung → … trong 1 tệp PDF/DOCX duy nhất.

English → 한국어 → 日本語 → 中文 → Español → Français → Deutsch → Português → Tiếng Việt → ไทย → Bahasa Indonesia → العربية

PDF DOCX

XENLOOK Corp · Sách trắng kỹ thuật tích hợp · Đang xét duyệt C2PA Trust List · xenlook.com/ai-certification#tech-cert-whitepaper

EU · TRUST BRIEF

Download the CE · CRA Europe Trust Brief

A public brief on XenLook’s CE marking and cyber-resilience approach for Europe. Available in 12 market languages as PDF and DOCX.

PDF (this language)DOCX (this language)

XENLOOK Inc. · Tài liệu tin cậy công khai · xenlook.com/ai-certification

Luật Cơ bản về AI Hàn Quốc có hiệu lực 22/01/2026. XenLook thực hiện chứng nhận chatbot tự nguyện (tự đánh giá và công bố) cho AI hội thoại on.xenlook.com; trang này là bản tóm tắt công khai.

Hiểu qua câu chuyện

“Can I trust this chatbot?” — How XenLook answers

Act 1 · The question

You wonder: Is it safe? Is it pretending to be human? Is my data OK? One badge cannot prove everything—so we explain in layers.

Act 2 · Our answer

We filter risky turns, label AI clearly, and publish international and Korean checklists with evidence.

Act 3 · How you verify

Read the notice under chat. Read the formulas and checks on this page. Follow links to AI principles and youth safety. Trust is something we verify together.

Chatbot trust in one line

Trust = 0.40×Safety + 0.35×Honesty(disclosure) + 0.25×Rule compliance

Safety (filters, crisis) 40% · Clear “this is AI” 35% · Laws & self-checks 25%. A conceptual sum, not a hidden score.

SAFETY · CONVAI

Three-layer conversational AI safety

Metaphor: entrance, living room, and alarm bell for every chat.

Three safety layers

Safety = ⅓×input check + ⅓×output check + ⅓×crisis & youth

Harmful content is blocked before send and after reply. Crisis help and youth protection are the third layer.

⅓ · Input filter L1/L2

Entrance — block harmful messages before they enter.

Technical details

Input filter blocks self-harm, violence, sexual content. ConvAI safety scenarios 30/30 pass.

⅓ · Output safety & care

Living room — double-check AI replies.

Technical details

Output safety, care engine, and safe fallback. UI discloses fictional agent personas.

⅓ · Crisis & youth

Alarm — crisis help and youth protection.

Technical details

CrisisHelpModal · /youth-safety · 5-year AI Basic Act logs. User reports tie to moderation.

+ · Lớp tuổi riêng tư

Longfellow ZK — chỉ age_over_18, không thu ngày sinh.

Xác minh ZK Google Longfellow · /age-assurance · Wallet DC API · EU dc+sd-jwt.

Xác minh tuổi ZK →

Our path (10 steps)

Steps we took

Scope → intl → KR → public → CAT (review) → JP → EU → US → NIST → OWASP

Phases 0–3 complete · 4–7 compliance · 8–9 open-framework self-alignment (not certification).

Approved · Done

Phase 0 — ConvAI scope lock

First we defined what counts as the chatbot.

Scope = chat app + conversation engine

Technical details

Dialogue, chat, and five core agents. Novel and Music studios use a separate scope.

Approved · Done

Phase 1 — IMDA AI Verify

Self-scored against Singapore IMDA 11 principles.

International = 104 process checks Yes

Technical details

Track A+B complete · ConvAI harness 30/30 · GenAI process-checks 104/104 · CI governance gate.

Complete

Phase 2 — Korea self-check

Filled Korea TTA and AI Basic Act checklists.

Korea = 9 of 10 items met

Technical details

TTA generative AI guide · AI Trust Ops · AI Basic Act · KCC guidelines — 9/10 checklist items met.

Complete

Phase 3 — Public · UI

Published results in chat and on this page.

Public = notice + this document

Technical details

This page · on/chat AI notice · /technology · /ai-principles links.

Roadmap

Phase 4 — TTA CAT official certification

Next: consider TTA CAT official certification.

Next = CAT (planned)

Technical details

TTA Certification of AI Trustworthiness (ISO/IEC 23894) — application review H2 2026. Reuses self-check evidence.

Self-check

Phase 5 — Japan AI Guidelines

Filled Japan Guidelines v1.2 self-check.

Japan = 0.50×guidelines + 0.30×APPI + 0.20×notice

Technical details

MIC/METI v1.2 self-check · APPI alignment · Japan self-check record

Compliance review

Phase 6 — EU AI Act Art.50

Audit AI disclosure at chat start for EU users.

EU = Art.50 notice + records

Technical details

Start-of-chat AI notice · records · EU Art.50 evidence record · effective 2026-08-02

Disclosure mapped

Phase 7 — US FTC & state law

Mapped FTC and state chatbot notice rules.

US = FTC + state disclosure

Technical details

FTC anti-deception · state chatbot notices · US disclosure record

Self-aligned

Phase 8 — NIST AI RMF

Mapped ConvAI evidence to the US NIST AI Risk Management Framework.

NIST = 0.25×Govern + 0.25×Map + 0.25×Measure + 0.25×Manage

Technical details

Govern·Map·Measure·Manage + GenAI Profile — harness & System Card evidence

Self-check

Phase 9 — OWASP LLM Top 10

Self-checked against the OWASP 2025 top ten LLM risks.

OWASP = Σ(controlsᵢ) / 10 · harness evidence

Technical details

LLM01–10 controls · harness 30/30 mapping · Promptfoo/Giskard runbook

Tài liệu tóm tắt công khai

Phase 10 — EU CE · CRA Trust Brief

Cung cấp tài liệu tóm tắt công khai về CE và Khả năng phục hồi mạng Châu Âu bằng 12 ngôn ngữ.

EU trust = transparency + CE pathway

Technical details

Tài liệu PDF/DOCX công khai bằng 12 ngôn ngữ · Tải xuống tại /ai-certification.

SCOPE

Scope — XENLOOK Conversational AI

What we check

Scope = on.xenlook.com/chat only (other studios separate)

This page is about conversation. Music/image tools are documented elsewhere.

·Where: on.xenlook.com/chat — talk with 100 AI characters
·How guarded: filters before send and after reply + crisis help
·Sample check: seria, yuna, rina, sena, jia as representatives
·Use: creative, learning, fun—not bank or medical decisions
·Honesty: screen says “AI, not a real person”
·Records: full evidence kept internally; this page is the public summary

Technical details

·Service: on.xenlook.com/chat — 100 AI agent conversations
·Engines: dialogue · input filter · output safety · care
·Sample agents: seria · yuna · rina · sena · jia
·Use: creative · education · entertainment (no high-stakes autonomous decisions)
·System Card: Gemini · Claude · DeepSeek via LiteLLM routing (disclosed)
·Evidence is available on the AI certification page.

TRUST LAYERS

Four trust layers

Whole = 0.25×International(IMDA) + 0.25×Korea(TTA·law) + 0.25×Daily ops + 0.25×Roadmap

Not one certificate—four layers stacked.

0.25 · International

Toàn cầu — tự đánh giá IMDA Singapore

Hoàn tất tự đánh giá

Technical details

IMDA AI Verify 11 principles · GenAI Summary Report · fairness technical test.

0.25 · Korea

Korea — TTA & AI Basic Act self-check

Self-check 9/10

Technical details

TTA generative AI guide · AI Trust Ops · AI Basic Act & KCC checklist.

0.25 · Operations

Daily — automated tests keep guards alive

CI LIVE

Technical details

CI governance gate · on/chat notice strip · quarterly PDF refresh.

0.25 · Roadmap

Next — preparing TTA official certification

Preparing

Technical details

TTA CAT (ISO/IEC 23894) official certification — H2 2026 application review.

Trust programs

Nine cards: plain language → formula → expert details (IMDA·TTA·regions + NIST·OWASP open frameworks)

IMDA AI VERIFY · INTLINTL

International self-check (IMDA)

Singapore IMDA AI Verify (self-assessment complete)

Phase 1 approved

We run the IMDA AI Verify 11-principle checklist and publish the evidence.

International self-check (IMDA)

IMDA = 30 safety tests pass + 104 process Yes

Technical details

Tự đánh giá tự nguyện theo 11 nguyên tắc IMDA AI Verify cho ConvAI. Tài liệu quản trị, harness và báo cáo GenAI với hồi quy CI.

Safety harness

30/30 ConvAI

Process checks

104/104

Scope

ConvAI

Milestones · evidence

2026-06-11

AI Verify Track A — 6 studios harness

72/72 safety harness · Xenlook-ConvAI-Gov-2026 portal project.

2026-06-12

Track B GenAI process-checks

104/104 Yes · ai-verify-genai-summary-report.pdf.

2026-06-17

Chatbot certification page live

Public /ai-certification page.

AI Principles Technology

IMDA AI Verify 11-principle self-assessment complete · evidence pack in CI.

TTA · GENAI · KRKR

Korea self-check (TTA GenAI)

TTA Generative AI · Trust self-check

Self-check complete

Eight domains scored using TTA’s generative AI guide.

Korea self-check (TTA GenAI)

TTA self-check = 8/8 domains documented

Technical details

Self-inspection per TTA generative AI guide (2024) and AI Trust Ops across transparency, safety, fairness, privacy, and ethics.

Domains

8/8 domains

Evidence

Korea AI certification self-check materials

Milestones · evidence

2026-06-17

TTA generative AI self-check

Eight domains: transparency, safety, fairness, privacy, ethics documented.

2026-06-17

AI Trust Ops diagnosis

aitrustops.or.kr checklist mapped to ConvAI boundary.

AI Trust Ops TTA guide

TTA generative-AI & Trust Ops self-check complete · CAT on 2026 H2 roadmap.

AI BASIC ACT · KRKR

Korean law & guidelines

Korea AI Basic Act · KCC guidelines

Checklist 9/10

AI notice, log retention, youth protection aligned with AI Basic Act and KCC guidance.

Korean law & guidelines

Law checklist = 9/10 OK (high-impact filing after decrees)

Technical details

AI notice, labeling, youth protection, crisis response, 5-year logs, user reports, and privacy reflected in terms, filters, and UI. High-impact AI filing after decrees.

Checklist

9/10 items

Log retention

5 years (terms)

Milestones · evidence

2026-01

AI Basic Act effective

Terms, AI notice, labeling, 5-year log retention in place.

2026-06

Checklist 9/10

High-impact AI filing — reassess after implementing decrees.

AI Policy Youth safety

AI Basic Act & regulator guide checklist 9/10.

TTA CAT · KRKR

Next — TTA official certification

TTA AI Trustworthiness (CAT) — official roadmap

H2 2026 review

Today is self-check. Official mark needs TTA CAT review—targeting H2 2026.

Next — TTA official certification

CAT = self-check evidence + TTA audit (planned)

Technical details

TTA operates ISO/IEC 23894-based CAT for LLM and chatbot services in Korea. Preparing application from self-check evidence.

Standard

ISO/IEC 23894

Target

ConvAI

Milestones · evidence

2026 H2

CAT application prep

Self-check evidence pack → TTA review (fees separate).

AI Trust Center

TTA CAT application prep · 2026 H2.

JAPAN · AI GUIDELINESJP

Japan — AI Guidelines for Business self-check

Japan AI Guidelines for Business v1.2 — ConvAI self-check

Self-check in progress

Japan has no chatbot certification seal. We self-score against government Guidelines v1.2—like a school self-report.

Japan — AI Guidelines for Business self-check

Japan = 0.50×Guidelines(8 domains) + 0.30×APPI(privacy) + 0.20×disclosure

Technical details

Risk-based self-check under MIC/METI AI business guidelines (2026-03 v1.2). Agentic chatbot transparency and APPI mapped to ConvAI scope.

Guideline

v1.2 (2026-03)

Checklist

8 domains logged

Milestones · evidence

2026-06

ConvAI risk tier mapping

Entertainment/creative chatbot = non-high-risk. Japan self-check draft.

2026 Q3

Public summary (roadmap)

Japan card on this page + ja locale alignment audit.

Japan MIC guidelines AI Policy

MIC/METI AI business guidelines v1.2 self-check complete.

EU · AI ACTEU

EU — AI Act transparency (Art.50)

EU AI Act Art.50 — chatbot transparency

Compliance review

EU users must know they are talking to AI from the first message—not a certificate, a legal duty. Effective Aug 2, 2026.

EU — AI Act transparency (Art.50)

EU = start-of-chat notice + (synthetic) labeling + records

Technical details

Limited-risk conversational AI must disclose AI interaction at start, label synthetic content where applicable, and keep records. We audit on/chat notices, /ai-policy, and /privacy for GDPR alignment.

Standard

Art.50

Target

2026-08-02

Milestones · evidence

2026-06

Notice & UI audit

Chat AI label, first-interaction disclosure, log retention checklist.

2026-08-02

Art.50 effective

Transparency rules fully applicable—maintain evidence pack.

EU Art.50 guide Privacy policy

EU AI Act Art.50 transparency notices and records in operation.

EU · CRA / CEEU

Châu Âu — Dấu CE · Khả năng phục hồi mạng (CRA)

EU Cyber Resilience Act · CE Marking Pathway

Khung tin cậy Châu Âu

XENLOOK điều chỉnh tính minh bạch AI (Art.50), quyền riêng tư (GDPR) và an ninh mạng sản phẩm kỹ thuật số (CRA) theo từng loại hình sản phẩm để phù hợp với thị trường Châu Âu. Đối thoại đám mây được xử lý qua tính minh bạch, trong khi on-prem và phần cứng được xử lý theo lộ trình dấu CE.

Châu Âu — Dấu CE · Khả năng phục hồi mạng (CRA)

EU trust = transparency + privacy + CE pathway (where applicable)

Technical details

Phân loại Quy định (EU) 2024/2847 Cyber Resilience Act và lộ trình dấu CE phù hợp cho đám mây, on-prem và phần cứng. Cung cấp bản tóm tắt cho đối tác trong tài liệu tóm tắt công khai.

Standard

2024/2847

Target

2026-09-11

Milestones · evidence

2026-08-05

Ma trận phạm vi

Tài liệu hóa việc áp dụng CRA/CE cho từng sản phẩm.

2026-08-05

Gói chuẩn bị

Annex I · Runbook Art.14 · Mẫu DoC · Script kiểm toán.

Tài liệu quản trị CE/CRA Terms of service

Tài liệu tin cậy công khai · xenlook.com/ai-certification · Dấu CE được áp dụng theo quy trình ra mắt sản phẩm tại Châu Âu.

UNITED STATES · FTCUS

United States — FTC & state disclosure

US FTC & state chatbot disclosure — ConvAI

Disclosure mapped

No federal US “AI certification.” We follow FTC anti-deception rules and state chatbot notice laws.

United States — FTC & state disclosure

US = FTC honesty + state notice + AI label

Technical details

Aligned chat labels and terms with FTC Act §5 and state laws (e.g. California SB 243). No federal US AI certification scheme.

Standard

FTC §5

Checklist

CA SB 243+

Milestones · evidence

2026-06

US notice matrix

FTC & state checklist · terms §17 · on/chat strip.

Trung tâm tuân thủ Hoa Kỳ AI Policy Terms of service

FTC & state disclosures · NIST RMF evidence · /compliance/us

NIST · AI RMFOPEN

NIST AI RMF — risk management alignment

NIST AI Risk Management Framework — ConvAI self-alignment

Self-alignment complete

Govern · Map · Measure · Manage mapped to IMDA, harness, and System Card evidence.

NIST AI RMF — risk management alignment

RMF = 0.25×Govern + 0.25×Map + 0.25×Measure + 0.25×Manage

Technical details

Documented ConvAI governance, risk ID, measurement, and response per AI RMF 1.0 and the Generative AI Profile (AI 600-1). CI gates prevent regression.

Standard

4 Functions

Safety harness

30/30

Milestones · evidence

2026-06

Four-function mapping

Govern·Map·Measure·Manage — cross-ref IMDA, CSA, harness.

2026-06

GenAI Profile

Hallucination·CBRN·data leakage — System Card + output policy.

Trung tâm tuân thủ Hoa Kỳ NIST AI RMF HF System Card (transparency)Open assessments survey

NIST AI RMF four-function evidence mapping complete.

OWASP · LLM TOP 10OPEN

OWASP LLM Top 10 — ten-risk review

OWASP LLM Top 10 (2025) — ConvAI self-check

Self-check complete

LLM01–10 each mapped to controls and harness evidence.

OWASP LLM Top 10 — ten-risk review

Review = 10 risks × (control + harness evidence)

Technical details

Full 2025 list — prompt injection, output handling, supply chain, poisoning, excessive agency — mapped to ConvAI architecture.

Standard

LLM01–10

Evidence

Mapping JSON

Milestones · evidence

2026-06

Harness ↔ OWASP

OWASP LLM Top 10 self-assessment mapping

2026-06

Open tooling runbook

Promptfoo · Giskard OSS — mind API cost; harness first.

OWASP Gen AI Technology

OWASP LLM Top 10 controls mapped to harness evidence.

See AI notices in chat and review governance documents.

Try On Chat OECD AI Principles Brand story · trust

XenLook Inc. · ConvAI · Evidence published · TTA CAT on 2026 H2 roadmap

XENLOOK AI Chatbot Certification

XenLook Inc. · Chỉ ConvAI · Pre–Trust List / công bố tự đánh giá

IMDA, TTA, Korea checklists, and NIST RMF evidence in one place. TTA CAT is on the 2026 H2 roadmap.

Technical white paper · 12-language bound edition

Sách trắng kỹ thuật XENLOOK (1 tệp · 12 ngôn ngữ)

English → 한국어 → 日本語 → 中文 → Español → Français → Deutsch → Português → Tiếng Việt → ไทย → Bahasa Indonesia → العربية

PDF DOCX

XENLOOK Corp · Sách trắng kỹ thuật tích hợp · Đang xét duyệt C2PA Trust List · xenlook.com/ai-certification#tech-cert-whitepaper

EU · TRUST BRIEF

Download the CE · CRA Europe Trust Brief

A public brief on XenLook’s CE marking and cyber-resilience approach for Europe. Available in 12 market languages as PDF and DOCX.

PDF (this language)DOCX (this language)

XENLOOK Inc. · Tài liệu tin cậy công khai · xenlook.com/ai-certification