AI Can No Longer Pretend to Be Human! EU's AI Act Imposes Fines up to €15 Million for Violations—AI Services from Taiwan Also Regulated

AI chatbots are becoming increasingly human-like, with many people unable to distinguish whether they are conversing with a real person or an AI. To prevent confusion, misinformation, and the misuse of generative AI such as deepfakes, the European Union has officially introduced the world’s strictest transparency regulations. Article 50 of the EU AI Act will take effect on August 2, 2026, across all 27 member states, mandating that AI must proactively disclose its identity and that AI-generated images, videos, audio, and text must carry identifiable markers. Violations could result in fines of up to €15 million or 3% of the previous year’s global annual revenue, whichever is higher.

The core principle of Article 50 of the EU AI Act can be summarized in one sentence: "AI must make it clear to users that they are interacting with artificial intelligence."

AI Must Disclose Identity from First Interaction—Chatbots Can No Longer Pretend to Be Human

From now on, any AI system that directly interacts with the public—including chatbots, virtual assistants, and autonomous AI agents—must clearly and recognizably inform users at the very first interaction that they are communicating with AI, rather than hiding such information in terms of service or usage guidelines.

According to the regulation, merely stating "This service is provided by AI" in the terms of service does not fulfill the disclosure obligation, as the law requires that the notification be completed at the first point of contact.

AI-Generated Images and Articles Must Carry Digital Watermarks

In addition to disclosing AI identity, another key requirement under Article 50 is that generative AI content must be identifiable. From now on, images, videos, audio, and text generated or significantly modified by AI must include machine-readable markers, enabling other systems to detect whether content was AI-generated or AI-altered, thereby reducing the risks of misinformation and deepfake dissemination. However, tools such as spell-checking or grammar correction that do not alter the substantive meaning of the original content are exempt from this requirement.

Three Transparency Obligations at a Glance

AI Interaction Disclosure: Chatbots, AI agents, and virtual assistants must inform users at first interaction that they are engaging with AI.

AI-Generated Content Labeling: Images, videos, audio, and text must include AI-generated markers that are machine-detectable.

Deepfake and Public Interest Disclosure: AI-generated deepfake content and AI-written texts on matters of public interest must disclose their AI origin. In cases where real individuals have reviewed and taken editorial responsibility, partial exemptions may apply.

Fines Up to €15 Million—Applies to Major Tech Companies Too

According to Article 99 of the AI Act, violations of Article 50’s transparency obligations can incur fines of up to €15 million or 3% of the previous fiscal year’s global annual revenue, whichever is higher. SMEs and startups may be subject to lower penalty thresholds to minimize impact on innovative companies. EU institutions themselves are capped at a maximum fine of €750,000. However, enforcement is not carried out directly by the European Commission but by the market surveillance authorities of each member state.

Taiwanese AI Companies May Also Be Affected

Many Asian operators mistakenly believe that as long as their company is not based in Europe, they do not need to comply with the AI Act. In fact, Article 50 follows a market-location principle—the determining factor is not the company’s place of establishment, but whether the AI service is offered to users within the EU. This means that even if a company is located in Taiwan, Japan, or the United States, if its AI products, chatbots, or generative content services are accessible to EU users, it may be required to comply with these transparency regulations.

Global AI Regulation Enters a New Phase

The European Commission states that the purpose of transparency regulations is to ensure people can clearly recognize when they are interacting with AI and to enhance the traceability and credibility of AI-generated content. Following the rapid proliferation of generative AI, the EU has formally elevated the principle of "AI must disclose its identity" from industry self-regulation to a legally binding requirement. With Article 50 taking effect on August 2, 2026, transparency will become the fundamental compliance threshold for any global AI service provider aiming to maintain a presence in the European market.