Responsibility statement

Multilingual media. Fair. Safe. For everyone.

We help you make content accessible worldwide – legally compliant, barrier-free, and transparent. Everything is made & hosted in Europe. Technologies such as voice cloning are used only with verifiable consent and are clearly labeled in the player.

Award: Best Employer 2025

What we stand for

At alugha, responsibility in handling video and audio content is at the core of what we do. We understand the power of digital media – and the risks.

Our goal: a platform that increases reach and efficiency while providing an ethical, transparent framework.

What guides us: authenticity, accessibility, data protection, respect for people and rights. We combine the speed of AI with human quality assurance (“hybrid intelligence”) to turn technology into meaningful, responsible impact.

Our commitments – concise & binding

  • Real consent instead of grey areas
    Voice cloning only on request, only with documented consent, and always clearly labeled in the player.
  • Clear content standards & a fair process
    Report → review → decision – with the option to appeal. Zero tolerance for illegal or clearly harmful content.
  • Data protection & hosting in Europe
    We use European data centres and disclose which providers are used for which purpose.
  • Accessibility by design
    Subtitles/CC, multilingual audio, audio descriptions, and structured review steps for inclusive content.
  • Privacy & Europe-first approach
    Hosting in Europe, full transparency on service providers and data flows.
  • Learning & giving back
    We support internships, mentoring, corporate volunteering and share knowledge through alucation.
  • Hybrid intelligence
    AI for speed, humans for quality. We combine both with defined review checkpoints.
  • Award-winning employer
    Best Employer 2025 – because good products can only grow in a good environment.

What you gain

  • More reach, less effort: One video, many languages – without duplicate files.
  • Lower risk: Clear rules for consent, labeling, and moderation.
  • Better experience: Accessible content is consumed and recommended more often.
  • Short paths: European infrastructure, direct support.

Authenticity & Handling of deepfakes

We reject manipulative deepfakes. At the same time, synthetic media can have legitimate use cases (e.g., accessibility, efficient localization). Our rules:

  • Voice cloning only with verifiable consent.
  • Clear labeling of synthetic audio tracks in the player.
  • Verifiable consent logs and documented approvals.

This protects personality rights and trust without blocking innovation.

Content standards & Moderation

We keep alugha free from illegal and clearly harmful content. Our process is transparent and fair.

Categories:
A – illegal / acutely harmful (priority) · B – risk content · C – contested cases

SLA Baselines:
Initial response ≤ 8 hours · Decision ≤ 48 hours (A-cases prioritized)

Workflow:

  • Report (email)
  • Preliminary automated heuristics
  • Escalation to the Federal Criminal Police Office (BKA)
  • Learning (improving rules/workflows)

Scope: misuse of AI features, prompt flows, moderation signals, disclosure mechanics.

Accessibility & Plural media spaces

Accessibility is not an add-on — it’s our standard. With subtitles, multilingual audio, closed captions, and audiodescription, we ensure that content reaches more people, regardless of language, ability, or device.

By enabling true accessibility, we create plural media spaces where all voices can be heard, seen, and understood.

Double materiality

Our responsibility is guided by two questions:

  • What impact do we have on the world? (impact materiality)
    Access to education and language, accessibility, protection from manipulation, media plurality, youth participation.

  • What impacts us? (financial materiality)
    Regulation, downtime, data protection, reputational risks, operational security.

Each year, we prioritize key topics from both perspectives, visualize them in a topic heatmap, and define concrete measures – published throughout the year.

Society & Politics – Actively practicing responsibility

We do not shy away from political responsibility. We contribute to associations and expert groups advocating for a strong, innovation-friendly German and European AI landscape.

Our CEO, Bernd Korz, is a member of the Federal Expert Commission on Artificial Intelligence at the German Economic Council. There, we share practical insights from product development, data protection, and accessibility projects to help shape responsible AI frameworks.

alucation – Sharing knowledge, opening opportunities

alucation is our educational initiative: free, open, and inclusive, making learning accessible in many languages.

Concretely this means:

  • Open learning library with multilingual, curated content (own productions, partner contributions, CC/public-domain materials with clean attribution).
  • Multilingual didactics: short, accessible units for schools, vocational training, universities, and lifelong learning.
  • Easy participation: educators, learners, and creators can contribute, translate, or voice content – with clear license guidance.
  • Practical tools: produce, subtitle, translate, and publish multilingual content directly inside the alugha ecosystem.

What is our AI trained on?

Our AI features (e.g., transcription, translation, quality signals) are improved using curated educational content from alucation and legally usable sources.

Principle: only with clear legal basis, documented consent, and clean attribution.

Sources we use

  • alucation content (our own or partner-created)
  • Licensed materials (contract/license)
  • CC/public-domain content with proper attribution
  • Synthetic/augmented data (e.g., accent/format variations)

What we explicitly do NOT use

  • No private customer data without a legal basis or explicit consent
  • No confidential material from protected or non-public projects
  • No content with unclear origin or violated rights

How we ensure compliance

  • Consent & licensing checks
  • Tagging every dataset with source/license/use
  • Data minimization & PII protection
  • Full audit trail and traceability

Opt-in/Opt-out for customers

  • Your projects are excluded from training by default (opt-out by default)
  • Opt-in requires separate written agreement
  • Voice-cloning models require prior written consent of the person concerned

Why alucation works as a training base

  • Didactic quality: clear language, defined learning goals, consistent terminology
  • Multilinguality: real accents, registers, domain vocabularies
  • Licensing clarity: transparent origin and usage rights

Hybrid Intelligence – AI + Humans

We use AI to become faster – and humans to stay accurate.

Our review steps include:

  • Terminology guides & style checks
  • A11y checks (subtitles/CC, readability, listening comprehension)
  • Four-eyes approval for sensitive cases (always included in our Full Service)
  • Documented decisions for traceability

Transparency in our supply chain

Trust requires insight. We publish an up-to-date list of our infrastructure and SaaS partners with:

  • purpose/service
  • processor role
  • region & storage location (EU/EEA)
  • data categories
  • certifications (if applicable)

So you always know where and why data is processed.

Digital circular economy

“One master, many languages” instead of redundant uploads:

  • less storage, less traffic, less maintenance
  • better searchability and faster updates

“One master, many languages” instead of redundant uploads:

  • average storage savings per video: approx. 1.95 GB (around 78%)*
  • estimated traffic reduction: approx. 60–80%*

    * Based on 10 videos, each 5 minutes long, in 5 languages.

Environment & Resource efficiency

We want multilingual media to be fair and resource-efficient. That’s why we optimize energy usage, data volume, and run-times end-to-end – from infrastructure to player.

Our commitments

  • Renewable-first infrastructure: Our storage and compute workloads operate in European data centres powered by renewable electricity. Across our wider infrastructure, we prioritize partners with strong sustainability commitments and transparent energy practices. While not every delivery component can guarantee renewable energy usage at all times, renewables form the backbone of our platform operations.
  • Data minimization & deletion cycles: We store only what is necessary and delete automatically after defined periods.
  • Short data paths within Europe: EU hosting reduces network transit (fewer backbone hops) and remote caching. This saves energy and improves latency.
  • One video, many languages – not countless copies: One video track + multiple audio/subtitle tracks. Users only download what they need.
  • “Green” AI practice: Optimized training and inference, small purpose-built models, reuse over retrain, strict data minimization, batch/off-peak jobs, feature-level measurements.

Implementation

  • Energy: preference for DCs using renewable electricity, energy-efficient cooling, and provider-level transparency where available.
  • Data: standard deletion policies for raw files and logs, pseudonymization.
  • Delivery: EU-local CDN/caching, streaming only required tracks, avoiding redundant encodes.
  • AI: fine-tuning instead of full retraining, augmentations instead of massive raw datasets, model-size evaluation.

People, Diversity & Public good

  • Diversity (team & leadership):We promote diverse backgrounds and perspectives. Decisions rely on competence – fair and transparent.
  • Wellbeing: Healthy work conditions: flexible models, reasonable workload, prevention over firefighting.
  • Youth participation: Structured internships with learning goals, feedback, mentoring.
  • Corporate volunteering: Support for educational, climate, social, and community initiatives.
  • Knowledge transfer: We share workshops, materials, and best practices – openly where possible.
  • Social taxonomy & common good reporting: We map our impact to the EU Social Taxonomy and publish a compact, transparent common good score.

Practical examples

  • Authenticity: You want to clone a speaker – we verify consent, label the track, and you distribute safely.
  • Moderation: A file is reported – we review, justify the decision, and allow appeals.
  • Accessibility: One upload, many languages/subtitles – more reach, less redundancy.

Frequently asked questions (FAQ)

Do you use deepfake technology?
No. We reject manipulative deepfakes. Voice cloning only with consent and clear labeling.

How do I report problematic content?
Email us. Every report is reviewed, with feedback and appeal options.

How do you handle data?
We host in Europe and work with selected providers. You’ll find the current list above under “Transparency in Our Supply Chain”.

How do you ensure accessibility?
Subtitles/CC, audio description, and structured review steps are part of our standard workflows.

What data is used to train your AI?
No private customer data without legal basis. Clear licensing and disclosure principles apply.

Contact

Questions, feedback or a report?
responsibility@alugha.com – we’re here for you.