A language is not a code. It is not a system of symbols that maps inputs to outputs. It is not a dataset to be tokenized, embedded, and compressed into a vector. A language is the accumulated wisdom, poetry, grief, humor, law, medicine, prayer, protest, and philosophy of every generation that has spoken it. When Hazaragi disappears, a thousand years of Hazara oral history, proverbs, and communal memory disappear with it. When Nuristani languages fade, the pre-Islamic cultural heritage of the Hindu Kush — preserved only in those languages — vanishes from human knowledge. When Dari is flattened into a single dialect by AI models trained predominantly on Kabuli speech, the medical vocabulary of Herati physicians, the legal traditions of Balkhi jurists, and the literary heritage of Badakhshani poets are erased.
Ariana Nexus processes Afghan languages every day — translating them, interpreting them, annotating them, validating AI models that generate them, moderating content written in them, and training professionals to understand the cultures they carry. This gives Ariana Nexus both a unique capability and a profound responsibility: the responsibility to treat every Afghan language not as a commodity to be processed but as a living cultural heritage to be preserved, respected, and accurately represented.
Language Integrity & Ethics is the framework that governs how Ariana Nexus fulfills this responsibility — through ethical standards for language work, commitments to linguistic diversity and endangered language preservation, and red lines that ensure no Ariana Nexus service or deliverable contributes to the erosion, homogenization, or misrepresentation of any Afghan language.
Afghanistan is one of the most linguistically diverse countries in the world. Ariana Nexus actively supports or plans to support all Afghan languages, recognizing that each carries irreplaceable cultural knowledge:
Dari (Farsi) — Indo-Iranian. ~12 million National language; multiple dialects (Kabuli, Herati, Mazari, Badakhshani) Full coverage — all major dialects
Pashto — Indo-Iranian. ~12 million National language; multiple dialects (Kandahari, Nangarhari, Wardaki, Waziri) Full coverage — all major dialects
Hazaragi — Indo-Iranian. ~3 million Distinct dialect/language of the Hazara people; increasingly recognized as separate Full coverage — treated as distinct linguistic system
Uzbeki — Turkic. ~2.5 million Major minority language Active coverage
Turkmeni — Turkic. ~1.5 million Minority language Active coverage
Balochi — Indo-Iranian. ~500,000 Minority language; cross-border with Pakistan and Iran Active coverage
Pashayi — Indo-Iranian. ~400,000 Endangered; multiple varieties Coverage development; preservation priority
Nuristani languages (Kati, Kamkata-viri, Askunu, Waigali, Tregami, Prasun) — Indo-Iranian (Nuristani branch). ~100,000 combined Endangered; unique pre-Islamic linguistic heritage Coverage development; critical preservation priority
Pamiri languages (Shughni, Wakhi, Ishkashimi, Munji) — Indo-Iranian (Eastern Iranian). ~100,000 combined Endangered; spoken in remote Wakhan Corridor and Badakhshan Coverage development; preservation priority
Brahui — Dravidian. ~200,000 Minority; remarkable Dravidian isolate in an Indo-Iranian region Coverage development
Kyrgyz — Turkic. ~5,000 Critically endangered in Afghanistan; spoken in the Wakhan Corridor Preservation priority; documentation support
Aimaq — Indo-Iranian. ~800,000 Semi-nomadic community language; distinct from standard Dari Active coverage development
Every Afghan language on this list — from Dari with its twelve million speakers to Kyrgyz with its five thousand — carries cultural knowledge that cannot be accessed through any other language. A medical concept that exists in Hazaragi may have no equivalent in Kabuli Dari. A legal principle embedded in Pashto tribal law may be untranslatable into English without losing its cultural meaning. A Nuristani creation narrative preserved only in Kati represents a pre-Islamic worldview that exists nowhere else on earth.
When AI models treat “Afghan languages” as a single category, when translation services offer only “Dari” without dialect specification, when content moderation algorithms evaluate Hazaragi speech against Kabuli Dari standards — they do not just produce errors. They participate in linguistic erasure. Ariana Nexus refuses to participate in this erasure.
No Afghan language is superior to any other. Dari is not more important than Pashayi. Pashto is not more valuable than Nuristani. The number of speakers does not determine the dignity or worth of a language. Ariana Nexus treats every language with equal professional respect, equal quality standards, and equal commitment to accuracy.
A Herati Dari speaker is not speaking “incorrect” Kabuli Dari. A Nangarhari Pashto speaker is not speaking “non-standard” Kandahari Pashto. Hazaragi is not “broken” Dari. Dialects are legitimate, complete, and historically rooted linguistic systems that deserve the same precision and respect as any standardized variety. Ariana Nexus never “corrects” dialect features to match a standardized variety unless specifically requested by the client for a documented purpose — and the CCB reviews any such request for linguistic ethics implications.
Translation is not the mechanical substitution of words from one language to another. It is the transfer of meaning, context, emotion, and cultural knowledge across linguistic boundaries. A translation that is linguistically accurate but culturally empty — that strips a Pashto proverb of its cultural resonance, that renders a Dari medical concept in clinical English without conveying the patient’s cultural understanding of their condition — fails the standard of language integrity even if it passes every linguistic quality metric.
AI language models trained predominantly on one dialect of a language (typically the most-resourced, highest-population dialect) will treat that dialect as the “standard” and other dialects as deviations. This is linguistic imperialism encoded in statistics. Ariana Nexus’s AI validation, annotation, and training services are designed to preserve dialectal diversity — ensuring that AI models recognize, correctly process, and appropriately generate all Afghan dialects rather than collapsing them into a single standardized form.
Endangered languages — Nuristani, Pamiri, Kyrgyz, Pashayi — are not languages that can be deprioritized because of their small speaker populations. They are languages that require more attention, more careful handling, and more expert validation precisely because they are at risk of disappearing. Ariana Nexus commits to contributing to the documentation, preservation, and accurate representation of endangered Afghan languages through its services and partnerships.
Interpreters, translators, and linguists are not interchangeable production units. They are cultural professionals whose expertise is the product of years of education, lived experience, and professional development. Ariana Nexus treats language workers with the respect their expertise deserves — fair compensation, safe working conditions, professional development opportunities, and the right to refuse assignments that compromise their professional ethics.
When Ariana Nexus lacks coverage for a specific language or dialect, it discloses this limitation honestly rather than substituting a different dialect or providing a lower-quality output without disclosure. If a client needs Wakhi interpretation and Ariana Nexus does not have a qualified Wakhi interpreter available, the client is informed — not given a Dari interpreter who “might understand some Wakhi.”
The following are absolute ethical red lines — actions that Ariana Nexus will refuse to perform regardless of client request, commercial value, or deadline pressure:
Red Line 1: No Dialect Erasure Ariana Nexus will not produce translations, annotations, or AI training data that erase dialect distinctions by treating all varieties of a language as interchangeable. If a client asks for “Dari translation” without dialect specification, Ariana Nexus will ask which dialect the audience speaks — it will not default to Kabuli Dari and hope for the best.
Red Line 2: No Linguistic Misrepresentation Ariana Nexus will not represent a translation as being in a specific dialect when it was actually produced in a different dialect. If a deliverable is labeled “Herati Dari,” it was produced by a Herati Dari speaker — not by a Kabuli Dari speaker who made their best approximation.
Red Line 3: No AI Training Data That Homogenizes Ariana Nexus will not annotate or validate AI training data in a manner that teaches the model to treat dialectal variation as error. If a speech-to-text model is being trained on Dari, Ariana Nexus will insist that the training data include and correctly label all major dialects — not train exclusively on Kabuli Dari and treat Herati or Hazaragi speech as out-of-distribution noise.
Red Line 4: No Endangered Language Exploitation Ariana Nexus will not use endangered language data for purposes that do not benefit the language community. If an AI lab wants to use Nuristani language data to train a model, the engagement must include provisions for how the model’s capabilities will benefit the Nuristani community — not just extract value from their linguistic heritage.
Red Line 5: No Cultural Stripping Ariana Nexus will not produce translations that strip cultural content to produce a “cleaner” or more “universal” output. If a Pashto text contains a culturally embedded metaphor, the translation preserves the metaphor and provides cultural context — it does not replace the metaphor with a culturally neutral English equivalent unless the client specifically requests this and the CCB approves.
Red Line 6: No Language Weaponization Ariana Nexus will not provide language services for purposes that weaponize linguistic identity — including using dialect identification to profile, target, or discriminate against individuals based on their linguistic background. A system that identifies Hazaragi speakers for the purpose of sectarian targeting will never receive Ariana Nexus’s assistance.
Interpretation (Medical, Legal, Government): - Interpreter must be a native speaker of the specific dialect required. - Consecutive interpretation accuracy target: ≥98% semantic fidelity. - Simultaneous interpretation accuracy target: ≥95% semantic fidelity. - Cultural context preserved — not just words converted but meanings conveyed. - CCB cultural compliance review for high-risk encounters.
Translation (All Types): - Translator must be a native speaker of the target dialect. - Translation accuracy assessed against Cultural Compliance Scorecard. - Terminology verified against the CKB Terminology Database for the specific dialect. - Back-translation verification for clinical and legal documents (Tier 1 engagements). - CCB review mandatory before delivery.
AI Annotation and Validation: - Annotators must be native speakers of the language/dialect being annotated. - Dialectal variation correctly labeled — not normalized to a standard variety. - Cultural hallucination assessment (SCHA) applied to all Afghan-language AI outputs. - AI models validated for dialectal coverage — not just aggregate accuracy.
Content Moderation: - Moderators must understand the specific dialect, cultural context, and community norms relevant to the content. - Classification decisions account for dialectal meaning differences. - False positive analysis includes dialectal misinterpretation assessment.
Ariana Nexus commits to contributing to the preservation of endangered Afghan languages through its operational activities:
Documentation support: When Ariana Nexus encounters endangered language data through its engagements (with client authorization), the CCB evaluates opportunities to contribute to linguistic documentation — vocabulary lists, grammatical descriptions, cultural terminology — that support preservation efforts.
Academic partnership: Ariana Nexus partners with university linguistics departments and language preservation organizations to support research on endangered Afghan languages.
AI for preservation: Ariana Nexus advocates for AI language model development that includes endangered Afghan languages — not just the high-resource languages that are commercially viable. When validating AI models, the CCB specifically evaluates coverage of minority and endangered languages as a quality dimension.
Community benefit: Any work involving endangered language data is conducted with the principle that the language community should benefit from the work — through improved language technology, documentation, educational resources, or other community-identified benefits.
Speaker involvement: Endangered language work is conducted with the involvement of native speakers and community members — not by linguists who have studied the language academically but do not speak it.
UNESCO Atlas of World’s Languages in Danger — Endangered language identification and preservation. Aligned — endangered Afghan languages identified; preservation commitment documented
UNESCO Convention for Safeguarding Intangible Cultural Heritage — Protection of linguistic heritage. Aligned — language treated as intangible cultural heritage
UN Declaration on the Rights of Indigenous Peoples (Article 13) — Right to revitalize, use, and transmit languages. Aligned — support for all Afghan language communities
Universal Declaration of Linguistic Rights (Barcelona, 1996) — Equal dignity of all languages. Aligned — Principle 1: Every Language Has Equal Dignity
National CLAS Standards (HHS OMH) — Language access in culturally appropriate form. Aligned — dialect-specific language access; no dialect substitution
Section 1557 — Qualified interpreters in patient’s language. Aligned — dialect-specific interpreter matching
NAJIT Code of Ethics — Interpreter accuracy, fidelity, completeness. Aligned — semantic fidelity targets; cultural context preservation
ISO 17100:2015 — Translation quality requirements. Aligned — native-speaker requirement; quality verification
ISO 5060:2024 — Translation and interpreting — General requirements. Monitoring — new standard published 2024; evaluation for alignment underway
EU AI Act (Article 10) — Data representativeness for AI training. Aligned — dialectal diversity in training data; no homogenization
EU AI Act (Article 50) — Transparency for AI-generated content. Aligned — content authenticity labeling framework applied to all AI-assisted language outputs
NIST AI RMF (MAP) — AI context including affected communities. Aligned — linguistic communities identified as affected populations
ISO 639 — Language identification codes. Aligned — correct language/dialect identification in all deliverables
Endangered Languages Project (Google/First Peoples’ Cultural Council) — Endangered language documentation and technology. Monitoring — partnership evaluation for Afghan endangered languages
For healthcare systems: When your Afghan patients speak Herati Dari, they receive a Herati Dari interpreter — not a Kabuli Dari interpreter who approximates. When your consent forms are translated into Hazaragi, they are translated by a Hazaragi speaker who understands that Hazaragi medical vocabulary differs from standard Dari in ways that affect clinical understanding. Language integrity is patient safety.
For AI labs: When you train your models on Afghan language data annotated by Ariana Nexus, you receive data that preserves dialectal distinctions — data that teaches your model that Hazaragi is not “noisy Dari” but a distinct linguistic system. Our red line against AI training data that homogenizes ensures that your model serves all Afghan speakers, not just the majority dialect.
For government agencies: When you produce multilingual communications for Afghan communities, you receive translations in the specific dialect of the community you are reaching — not a one-size-fits-all “Dari” translation that connects with some communities and alienates others. Language integrity is effective governance.
For immigration attorneys: When your client’s asylum testimony is interpreted, the interpreter speaks your client’s dialect and understands the cultural weight of every word. A term that means “political activity” in one dialect may mean “armed resistance” in another. Language integrity is legal accuracy.
For the Afghan diaspora: Ariana Nexus exists to serve you — in your language, your dialect, your cultural context. We do not simplify your language for the convenience of AI models. We do not erase your dialect because a standardized version is easier to process. We do not treat your endangered language as less worthy of quality because fewer people speak it. Your language is your heritage, and we treat it accordingly.
If your organization requires language quality documentation, dialect-specific service capabilities, or a language integrity briefing, contact trust@ariananexus.com or +1 (202) 771-0224.
Current (2026) — Seven Language Ethics Principles codified; Six ethical red lines enforced; Full coverage of Dari (all dialects), Pashto (all dialects), and Hazaragi; Active coverage of Uzbeki, Turkmeni, Balochi, Aimaq; Coverage development for Pashayi, Nuristani, Pamiri, Brahui, Kyrgyz; Language quality standards per service type; CCB language integrity review integrated into all services; Endangered language preservation commitment. Operational
Hardening (Q3–Q4 2026) — Dialect-specific terminology database expansion; Endangered language partnership development; Language ethics training module for all language workers; Dialect coverage gap analysis and recruitment plan; Client-facing dialect specification guidance. In Planning
Scale (2027) — Full active coverage of all 13+ Afghan languages; Published Language Integrity Standard (public version); Academic partnerships for endangered language documentation; SOC 2 Type II evidence for language quality controls; Endangered language AI model validation capability. Planned
Certification (2028) — ISO 17100 alignment verification; ISO 42001 certification with language integrity controls; Third-party language quality audit; Language ethics contribution to international standards discussion. Planned
Expansion (2029–2030) — Multi-diaspora language integrity framework (Somali, Syrian, Rohingya, Ukrainian); Endangered language AI preservation initiative; Real-time dialect identification and routing; Language integrity API for client integration. Planned
Long-Horizon (2030+) — Global language integrity standard; AI-assisted endangered language documentation; Language preservation as a core AI safety principle; Language integrity architecture maintained through 2080 horizon. Vision
Linguistic Complexity. Languages are living systems that evolve continuously. Dialect boundaries are not always clearly defined, and native speakers may disagree on terminology, usage, and classification. Ariana Nexus applies professional judgment based on the best available linguistic expertise but does not warrant that its language assessments will satisfy every native speaker or linguist.
Coverage Limitations. Ariana Nexus actively supports or is developing coverage for all Afghan languages listed in this document. Some languages, particularly endangered languages with very small speaker populations, may have limited immediate availability. Ariana Nexus discloses coverage limitations honestly per Principle 7.
Endangered Language Preservation. Ariana Nexus’s commitment to endangered language preservation is implemented through its operational activities and partnerships. Ariana Nexus is not a linguistic preservation organization and does not conduct independent language documentation programs.
Limitation of Liability. TO THE MAXIMUM EXTENT PERMITTED BY APPLICABLE LAW, ARIANA NEXUS’S TOTAL AGGREGATE LIABILITY FOR ALL CLAIMS ARISING OUT OF OR RELATED TO LANGUAGE INTEGRITY, TRANSLATION ACCURACY, INTERPRETATION QUALITY, OR DIALECT SPECIFICATION SHALL NOT EXCEED THE AMOUNTS SET FORTH IN THE APPLICABLE ENGAGEMENT AGREEMENT, OR, WHERE NO ENGAGEMENT AGREEMENT EXISTS, ONE HUNDRED DOLLARS ($100). ARIANA NEXUS SHALL NOT BE LIABLE FOR ANY INDIRECT, INCIDENTAL, SPECIAL, CONSEQUENTIAL, PUNITIVE, OR EXEMPLARY DAMAGES ARISING FROM OR RELATED TO LANGUAGE SERVICES, DIALECT ACCURACY, OR LINGUISTIC ASSESSMENTS. NOTHING IN THIS SECTION SHALL LIMIT OR EXCLUDE ARIANA NEXUS’S LIABILITY FOR: (A) FRAUD OR FRAUDULENT MISREPRESENTATION; (B) DEATH OR PERSONAL INJURY CAUSED BY NEGLIGENCE; OR (C) ANY OTHER LIABILITY THAT CANNOT BE EXCLUDED OR LIMITED BY APPLICABLE LAW, INCLUDING BUT NOT LIMITED TO LIABILITY UNDER THE UK UNFAIR CONTRACT TERMS ACT 1977, THE UK CONSUMER RIGHTS ACT 2015, OR GDPR.
Dispute Resolution. Any dispute arising out of or relating to this page shall be subject to the dispute resolution provisions in the Terms of Use, Section 18.
This page is provided for informational purposes and does not constitute a warranty, guarantee, or binding commitment regarding Ariana Nexus’s language coverage or linguistic accuracy. Language services are subject to availability. Nothing in this page shall be construed as a waiver of any right, defense, or immunity available to Ariana Nexus under applicable law.
