距离欧盟人工智能法案生效还有 28 天
了解更多 →

欧盟人工智能法案透明度义务:第13条和第50条

欧盟人工智能法案将透明度分为两种制度。第13条规定了高风险人工智能的提供者和部署者之间的关系。第50条规定了人工智能系统与他们互动的人之间的关系——涵盖聊天机器人、深度伪造和人工智能生成的内容。

第13条和第50条

欧盟人工智能法案透明度义务:第13条和第50条

欧盟人工智能法案将透明度分为两种制度。第13条规定了高风险人工智能的提供者和部署者之间的关系。第50条规定了人工智能系统与他们互动的人之间的关系——涵盖聊天机器人、深度伪造和人工智能生成的内容。

上次更新: 2026年7月4日

第13条:对部署者的透明度

高风险人工智能的提供者必须向部署者提供足够的信息,以便正确使用该系统并进行有效的监督。

  • 提供者的身份和联系方式
  • 预期用途、准确性、稳健性和网络安全属性
  • 已知限制和已知可预见的滥用
  • 对特定人员或群体的表现
  • 输入数据要求和预期的输出解释
  • 人工监督措施以及如何实施这些措施
  • 维护和生命周期预期

第50条:对用户和公众的透明度

第50条的适用范围比仅高风险系统更广泛。四个核心义务:

  • 人工智能互动披露——必须告知自然人他们正在与人工智能系统互动(例如,聊天机器人),除非从上下文中可以明显看出。
  • 合成内容标记——人工智能生成的声音、图像、视频或文本必须是机器可读的,并且可以检测出是人工生成的。
  • 深度伪造披露——描绘真实人物或事件的内容必须被披露为人工生成,但明确的艺术、讽刺或虚构作品除外。
  • 公共利益文本——人工智能生成的内容如果用于向公众提供有关公共利益相关事宜的信息,则必须进行披露,除非有进行人工审查和编辑责任。

第13条和第50条如何相互作用

这两个透明度制度在价值链的不同层级上运行:

第13条是提供者和部署者之间的B2B义务。第50条面向最终用户。高风险聊天机器人的提供者既要向其企业部署者提供第13条规定的文档,又要向与聊天机器人互动的自然人提供第50条规定的信息。

Article 50 Exceptions and Implementation Checklist

Article 50 disclosure duties come with narrow exceptions, and Article 13 documentation has concrete build requirements. Work through both:

  • Interaction disclosure exception — you can skip the "you are talking to an AI" notice only where it is obvious to a reasonably well-informed person from the context of use; when in doubt, disclose
  • Deepfake artistic exception — for evidently artistic, creative, satirical, or fictional work, the disclosure can be limited to a form that does not hamper the display or enjoyment of the work, rather than omitted entirely
  • Law-enforcement exception — detection, prevention, investigation, and prosecution of criminal offences may be exempt from certain synthetic-content and interaction duties where authorised by law
  • Machine-readable marking — synthetic-content marking must be effective, interoperable, robust, and reliable as far as technically feasible; plan for a durable technique such as watermarks, metadata, or cryptographic provenance signals
  • Article 13 instructions for use — package provider identity, intended purpose, accuracy and robustness metrics, known limitations, and human-oversight measures into deployer-facing documentation you can version and update
  • Timing — Article 50 transparency duties apply from 2 August 2026; build the disclosure and marking mechanisms into your product now rather than retrofitting them

This checklist is an editorial summary, not legal advice; confirm the exact conditions of each exception against the regulation text and any AI Office guidance before relying on it.

How AIAgentree helps

AIAgentree gives you the disclosure records and deployer-facing evidence that Articles 13 and 50 expect, without changing how your agents run.

  • Tamper-evident decision records capture when an AI interaction, synthetic-content, or deepfake disclosure was shown, so you can demonstrate Article 50 compliance to an auditor after the fact
  • Human-oversight and approval workflows document the oversight measures Article 13 requires deployers to be able to exercise, with a record of who approved or overrode a decision and why
  • Structured exports over REST, MCP, A2A, and OpenTelemetry — using the Python and TypeScript SDKs — let you fold decision evidence into the instructions-for-use and technical documentation you owe deployers, hosted with EU data residency in Germany

Frequently Asked Questions

What is the difference between Article 13 and Article 50 transparency?

Article 13 is a provider-to-deployer duty: providers of high-risk AI must give deployers instructions for use covering purpose, accuracy, limitations, and human-oversight measures. Article 50 is a system-to-person duty: people must be told when they are interacting with AI or seeing AI-generated or manipulated content. A single high-risk chatbot can owe both.

When do the Article 50 transparency obligations apply?

The Article 50 transparency obligations for providers and deployers apply from 2 August 2026 under Regulation (EU) 2024/1689. Building disclosure and synthetic-content marking into a product early is far cheaper than retrofitting it after launch.

Do I always have to label AI-generated content?

Not in every case. Article 50 provides narrow exceptions — for example, where AI interaction is obvious from context, for evidently artistic or satirical works (where a limited-form disclosure suffices), and for certain law-enforcement uses authorised by law. Outside those conditions, synthetic and deepfake content must be marked as artificially generated or manipulated.

Does the deepfake disclosure exception mean I can skip marking artistic content?

No. For evidently artistic, creative, satirical, or fictional work the disclosure may be presented in a way that does not hamper the display or enjoyment of the work, but it must still be disclosed. The exception limits the form of disclosure, it does not remove it.

Continue exploring the EU AI Act guide

EU AI Act Compliance Guide

The complete guide to EU AI Act compliance for AI agents — start here.

Article 12 — Record-Keeping & Logging

What every high-risk AI system must log, and how to capture it.

Article 14 — Human Oversight

Designing effective human-in-the-loop controls for AI decisions.

Annex III — High-Risk AI Systems

Which AI use cases the Act classifies as high-risk.

EU AI Act Compliance Checklist

A step-by-step checklist to reach and document compliance.

Compliance Cost Calculator

Estimate your EU AI Act compliance effort and cost.

Deadlines & Timeline

Key enforcement dates, including the August 2, 2026 deadline.

Fines & Penalties

Penalty tiers up to €35M or 7% of global annual turnover.

Risk Management & Conformity Assessment

Build a risk management system and assess conformity.

GPAI Obligations

Rules for providers of general-purpose AI models.

EU AI Act for US Companies

Extraterritorial scope and what US providers must do.

Omnibus Update

The latest changes to the EU AI Act timeline and rules.

Penalty Calculator

Estimate your maximum fine under the Article 99 tiers.

Article 11 + Annex IV

What technical documentation the EU AI Act requires.

Article 26: Deployer Obligations

What deployers of high-risk AI must do, including log retention.

Article 17: Quality Management

The QMS providers of high-risk AI must document.

Article 10: Data Governance

Data quality, bias mitigation, and governance duties.

Article 4: AI Literacy

The staff AI-literacy duty in force since February 2025.

Deployer vs Provider

Who bears which obligation — and when a deployer becomes a provider.

FRIA (Article 27)

Who must run a Fundamental Rights Impact Assessment, and how.

Who Does It Apply To?

Scope, operators, and the extraterritorial reach of the EU AI Act.

Post-Market Monitoring

Articles 72–73: ongoing monitoring and incident reporting.

ISO 42001 vs EU AI Act

How the voluntary standard and the binding law fit together.

NIST AI RMF vs EU AI Act

A practical crosswalk between the framework and the law.

EU AI Act for Healthcare

High-risk medical AI, MDR/IVDR interplay, and clinician oversight.

EU AI Act for Financial Services

Credit scoring, insurance pricing, and existing financial regulation.

EU AI Act for HR & Employment

Hiring AI as high-risk, plus NYC LL144 and EEOC overlap.