Stay compliant with the EU AI Act

The AI Act is enforceable now, and every model version restarts the clock. LILT keeps your disclosures, docs, and training compliant in every market language.

The EU AI Act explained

What, who, when, and what it costs you.

The EU AI Act sets obligations on anyone who puts an AI system on the EU market or deploys one inside the EU. Language requirements come in two ways. Some duties name a language: your declaration of conformity is translated for the authorities of every market you sell into, and your technical file goes to a national authority, on request, in the language it names. Others name an audience: instructions have to be understood by your deployers, disclosures by the person at first interaction. Either way, your market list decides your language list.

High-risk systems defined under the act carry the documentation load, covering biometrics, critical infrastructure, education, hiring and workforce management, credit and insurance, law enforcement, migration, and justice. Chatbots and synthetic media carry transparency duties.

Enforceable since 2 August 2026. High-risk documentation obligations under Article 6(1) apply from 2 August 2027. Penalties reach EUR 15 million or 3% of global turnover.

What are the language requirements under the EU AI Act?

Instructions for use, in the languages your deployers read

Instructions behind a high-risk system, in the language of each deployer's member state. Re-issued every time the model version changes. (Art. 13; Recital 72)

EU Declaration of conformity, translated for each market

Into a language the national authorities understand, in every market where you make the system available. (Art. 47)

Technical documentation, in whichever language is asked for

Available in one language until an authority requests it. Then it is due in the language that authority names, on that authority's clock. (Art. 11, Art. 21)

Chatbot and interactive disclosures

Anyone interacting with your AI needs a clear notice at first contact. (Art. 50(1), 50(5))

Content marking and deepfake disclosures

Providers mark synthetic audio, image, video and text in a machine-readable format, detectable as artificially generated. Deployers disclose deep fakes clearly to the people exposed to them, at first exposure. (Art. 50(2), 50(4), 50(5))

AI literacy training

Staff and anyone operating a system on your behalf need a sufficient level of AI literacy. (Art. 4)

One Secure Enterprise-grade Multilingual AI Platform combined with Expert Services

Centralized and streamlined workflows powered by AI, agentic automation, and human experts worldwide. Reduce the cost and complexity of meeting AI Act requirements today and manage language requirements at scale for the long term.

Instructions for use to declaration of conformity, in any EU language

Multilingual AI models customized to your organization's terminology translate accurately and at speed across EU languages, with human expert verification where needed. With the largest ecosystem of 100+ pre-built connectors, LILT plugs into existing tools your content already lives in and seamlessly translate.

Disclosure strings, inside the release, in every EU market you serve

Translate disclosure text right where they live, in your code repositories, resource files and design tools. Placeholders, character limits and do-not-translate rules are enforced without a round trip.

AI literacy training, in languages as needed

Localize the training modules, e-learning materials, assessments and attestation records your workforce needs, with voiceover, subtitles and on-screen text handled in one workflow. Course files go back into your LMS ready to publish, refreshed as the curriculum changes.

One centralized platform for every regulated document

All documentation translated and the approval records behind them live in one enterprise-grade secure platform, with terminology, memory, version history. Each version reuses the last one's approved language, so you pay for the delta, not the document.

Human review where the exposure is

A disclosure string and a high-risk technical file do not carry the same risk. You set the tier. Instant AI translation and AI Review agents for low-risk content, expert-verified for anything that is high-stakes.

Support that covers you around the clock

Round-the-clock coverage across regions and time zones, with a dedicated support team that knows your terminology and your workflows and stands by on Slack, email, or call to help right away, not next business day to help you meet compliance.

One platform, not a patchwork

Regulated, in-product and training content at scale

Various vendors, each briefed separately and queued separately

Disclosure strings, UI text, e-learning modules, voiceover and subtitles in one connected workflow

Keeping up with content updates

A new quote and a new queue every time the model changes

100+ connectors into the tools your content already lives in, so documentation ships faster

Training and e-learning

A separate e-learning vendor briefed and queued after the source is final

Modules, voiceover, subtitles, assessments and on-screen text in one workflow, returned ready for your LMS

Audit trail and records

Evidence reassembled from email threads and vendor invoices

End-to-end governance with audit trails and quality control across every language version

Risk-tiering

One rate for everything, so a help-centre article costs what a safety label costs

You set the tier per content type, from instant AI for low risk to expert-verified for anything a regulator reads

Cost over time

Full price for the same work again, and the same terminology argument every round

Adaptive AI where every verifier edit retrains your model in real time, so each project needs less review than the last

Frequently asked questions

Who does the EU AI Act apply to?

Anyone who places an AI system on the EU market or deploys one inside the EU, including organisations based outside the EU. Providers and deployers carry different duties, and the obligations can reach systems already in use: a system on the market before 2 August 2026 is caught once it undergoes a significant design change (Art. 111(2)), not only new releases.

What does the EU AI Act require you to translate?

Instructions for use and technical documentation for high-risk systems, and the EU declaration of conformity. Where a duty is written for the user, it lands in the language of the member state where the system is used. Disclosures are required for chatbot and interactive disclosure strings, AI content labels and marking text, and deepfake and synthetic-content notices.

When did the EU AI Act take effect?

It has been enforceable since 2 August 2026. The obligations covering high-risk instructions and technical documentation land on 2 August 2027.

What are the penalties under the EU AI Act?

For the duties covered on this page, penalties reach EUR 15 million or 3% of global turnover. The Act sets several tiers, with the highest reserved for prohibited practices, so the exposure depends on which obligation you miss.

Do AI chatbot disclosures need to be translated?

Article 50(1) requires that people are told they are interacting with an AI, but it does not carry an explicit translation mandate. In practice the notice has to be understood by the person at first interaction, so teams serving multiple markets localize it.

Does the AI Act apply to AI systems we already have in production?

It depends. Under Article 111(2), a high-risk system you placed on the market before 2 August 2026 only falls under the Act if you make a significant change to its design after that date, unless it is used by a public authority, in which case compliance is required by 2 August 2030 regardless of any change. Most teams are caught because model updates are frequent, not because every system in the field is automatically covered.

How often does AI Act documentation need to be re-issued?

Article 13 instructions and technical documentation have to reflect the system as it currently is, so a material model change means a new version of the documentation in every required language. This is what makes the Act a continuous obligation rather than a one-off project.

What is Article 4 AI literacy training?

Article 4 requires providers and deployers to ensure their staff have a sufficient level of AI literacy. In practice that means training material every affected employee can read, which for a multinational workforce may involve producing it in every working language.

EU AI Act Requirements At A Glance

Content / document	Translation requirement?
1. Instructions for use for high-risk AI systems	YES, explicit. They must be understandable to the target deployers in the language determined by the Member State.
2. EU Declaration of Conformity	YES, explicit. It must be translated into a language easily understood by the competent authorities in the Member States where the high-risk system is marketed or made available.
3. Technical documentation and compliance documentation requested by authorities	YES, when requested. Providers must supply the necessary information and documentation in an EU official language indicated by the Member State authority. Not necessarily translated into every market language in advance.
4. Chatbot and AI interaction disclosure	Disclosure required. Potential localization, but not an explicit translation mandate. Users must be informed they are interacting with AI. Article 50 does not say "translate into the language of every interface."
5. Emotion-recognition and biometric categorisation notices	Disclosure is required, but the Act does not explicitly prescribe translation into each market language.
6. Deepfake disclosure notices	Disclosure required. Localization may be necessary to make it meaningful, but there is no explicit language-of-market translation clause.
7. AI-generated public-interest text disclosures	Disclosure is required, not expressly a translation requirement.
8. AI literacy and training materials	NO explicit translation requirement. Potentially an organisation's localization call into every employee's working language to ensure literacy.