More Real Than Real — How Can We Tell What's Fake?
"Don't Be Fooled" – Hyper-Realistic AI Videos Are Now Being Used in Public Campaigns
A recent video went viral on social media, capturing widespread attention. Framed as a breaking news broadcast, it opens with a shocking report: "Lava is erupting in downtown Seoul."
Throughout the video, various anchors, reporters, students, and celebrities appear — but at the end, each reveals:
"I'm AI. Don't fall for it."
The video highlights how incredibly difficult it has become to distinguish between real and AI-generated content.
This campaign was created using Google’s video-generating AI model Veo as part of a public awareness initiative showcasing the risks of advanced AI content. Many viewers reacted with alarm: "It’s scary how real it looks." and "I don’t know what’s real anymore."
How Did AI Content Become So Hyper-Realistic?
AI is not a “creator” in the way we typically imagine. Instead, after learning from vast amounts of data, it predicts plausible outcomes by combining the most probable words, pixels, and sound fragments.
That’s why AI writing models follow the natural flow of language, AI image generators combine facial proportions and lighting, and AI video models add realistic facial expressions and voices.
The results are highly natural — sometimes so real that they appear even more authentic than reality itself, leading people to mistake them for the real thing.
When Lines Blur — The Threats to Security and Trust
AI-generated content, once seen as merely fascinating technology, is now being used as a tool that can cause real harm and social disruption.
- Impersonating celebrities using AI deepfake videos
- Voice phishing scams mimicking family members with AI-generated voices
- Manipulating public opinion through AI-generated news content
In today’s environment where fakes can circulate as if they were real, trusting content simply because it "looks convincing" is becoming increasingly dangerous.
How Are Our Society and Governments Responding?
As AI-generated content becomes so sophisticated that it is nearly indistinguishable from reality,
the issue has evolved into a social challenge that can no longer be managed by individual discernment alone.
Governments and global platforms are now establishing institutional frameworks to address digital disruption.
🌐 Global Regulatory Developments
- European Union
- Under the AI Act, generative AI content must carry origin labels ("AI-generated") to ensure transparency.
- United States
- Social media and streaming platforms are under pressure to implement automatic AI content tagging systems, while discussions are underway to classify manipulated video and audio as “modulated content” and to specify legal accountability.
- China
- The “Interim Measures for the Administration of Generative AI Services” were jointly announced on July 13, 2023, by seven government agencies, including the Cyberspace Administration, Ministry of Education, and Ministry of Science and Technology, and came into effect on August 15 of the same year.
- The provisional regulation mandates identification markings for outputs generated by generative AI.
🇰🇷 South Korea’s Response
- Korea AI Basic Act (scheduled to take effect in 2026)
- Generative AI outputs—including movies, dramas, and YouTube content—will be required to carry an "AI-generated" label.
- Ministry of Science and ICT
- Developing standard technical guidelines for AI content identification,
- and collaborating with platforms to conduct pilot projects for the deployment of detection technologies.
- Seoul Metropolitan Office of Education
- Preventive measures: AI ethics education introduced at the elementary level
- Since 2022, the Seoul Metropolitan Office of Education has been developing and distributing AI ethics education materials for elementary school students.
- These programs aim to cultivate critical thinking skills from an early age, teaching children not to blindly trust AI.
So, How Can the General Public Distinguish AI Content Today?
AI-generated content has become so sophisticated that it is now virtually impossible for the human eye alone to tell real from fake.
In the past, people could sometimes spot clues—such as unnatural fingers, blurry text, or distorted proportions—but now even these traces are increasingly difficult to find.
That’s why the first mindset we need is to develop a healthy habit of asking, “Is this real?” And when necessary, we can also make use of the following tools to check whether content was AI-generated.
Leading AI Detection Tools Available Today
✅ 1. AI or Not
- Analyzes images and text to detect AI generation
- Free for public use → aiornot.com
✅ 2. Hive AI
- Content moderation platform (image/text/voice detection)
- Used for corporate advertising and social media content analysis → thehive.ai
✅ 3. FakeCatcher (Intel)
- Deepfake detection based on skin blood flow analysis
- Being integrated into various platforms → Introduced by Intel
✅ 4. GPTZero / Originality.ai
- Detects AI-generated text
- Commonly used by bloggers, editors, and educational institutions
Note: The AI detection market is still technologically immature. Many tools are business-oriented, and their accuracy is not 100% guaranteed.
In other words, these tools serve only as reference points—ultimately, the most important thing is to develop the habit of questioning and verifying.
What We Need Now: The Habit of Verification
AI does not provide us with the truth.
That’s why we must avoid accepting all content at face value. Instead, we need to check sources, consider context, and maintain a habit of healthy skepticism.
This is our most fundamental defense and essential survival skill in the age of AI.
📌 Coming Up Next
So, who legally owns AI-generated content?
In the next installment, we’ll explore copyright issues and the potential uses of AI-generated works.