Hangzhou, Zhejiang, China – September 10, 2026
AI-Generated Video Output 50× Higher Than Human? Regulatory Framework Takes Effect – And Infrared Filters Become the Hidden Line of Defense
Recently, an unexpected trend has emerged on short-video platforms: AI-generated videos now outnumber human-produced ones by a factor of 50, yet they account for only 1/25 of total views. In other words, the vast majority of AI-generated content is never actually seen by ordinary users. It is being mass-produced and automatically distributed in the background at extremely low cost and high speed, forming a vast “dark content pool.”
In the short-drama space, the numbers are even more striking—data shows that over 95% of micro-short-drama content is now either partially or fully AI-generated. From scriptwriting to storyboarding to post-production, the traditional film and television industry chain is being fundamentally restructured.
Production capacity is not the problem. The problem is whether viewers know what they are watching.
The Government Has Stepped In
Around July this year, China’s National Radio and Television Administration (NRTA) and other regulatory bodies pushed forward a major development: the mandatory implementation of AI content labeling regulations has entered its rigid enforcement phase. In short, what was previously a voluntary, platform-self-regulation guideline for labeling AIGC (AI-Generated Content) has now become a mandatory standard—failure to label may result in penalties.
The core logic of this regulatory framework is clear: as AI-generated content becomes increasingly realistic and less expensive to produce, allowing it to circulate freely without distinction directly undermines public trust in the authenticity of information. You don’t know whether the news footage you’re watching is synthetic. You don’t know whether the voice you’re hearing is an AI voice clone. Even the facial image you’re looking at could have been replaced by a deepfake.
Thus, the labeling system is not a technical detail—it is the foundational infrastructure of digital content ecosystem governance.
Technical Challenges in Compliance Implementation
Labeling itself is not difficult—watermarking works. But building a watermarking system that is traceable, verifiable, and tamper-resistant requires a complete technical infrastructure. Three layers of capability are worth breaking down.
Layer 1: Digital Watermark Embedding. Invisible digital identifiers are embedded into the metadata of AI-generated content, and these identifiers can be read and verified through specific algorithms. Robustness is the core metric—the watermark must remain extractable even after the video undergoes compression, cropping, color grading, and other common processing. Good solutions embed watermark signals at multiple modalities during the AI training phase, not just as surface-level marks on the final output layer.
Layer 2: Multi-Spectral Anti-Counterfeiting and Traceability. Most people haven’t yet recognized the importance of this layer—yet it is arguably the most hardcore technical defense line. The reason multi-spectral liveness detection holds up in the authenticity battle is precisely because it relies on differences in reflectance characteristics across different wavelengths. This physical-layer authentication capability is unforgeable. Extending this approach to content security leads to a key insight: the real world’s complex light-shadow relationships contain coupled information across multiple bands—visible, near-infrared, and thermal infrared. No AI synthesis process can accurately simulate the spectral response patterns across all bands simultaneously.
This means that if front-end capture devices are equipped with multi-spectral sensors (for example, cameras fitted with near-infrared narrowband filters), they can capture authenticity evidence beyond the visible spectrum. AI-generated images may be perfect in the RGB domain, but in the reflectance distribution of the near-infrared band, there is an unbridgeable physical gap between screen pixels, printing toner, and real human skin. MULTI IR’s 850nm/940nm narrowband filters are critical components in such applications—their passband control precision and out-of-band rejection directly determine the data quality available to downstream anti-counterfeiting algorithms.
Layer 3: Full-Chain Content Auditing and Archiving. From the capture device (original files containing spectral calibration data), through transmission channels (blockchain audit nodes), to the publishing end (audio/video files with digital signatures)—every step has hash values and audit logs. Once content is tampered with, the signature verification fails.
These three layers together form a complete content trust system covering the entire journey from acquisition to distribution.
Industry Shifts in the Compliance Era
Let’s start with the most direct impact:
The cost advantage for small and individual creators may be compressed. One of AI-generated content’s biggest competitive advantages has been its extremely low barrier to entry—any ordinary user can generate a video with a few prompts. But if all content now requires compliance labeling and subjection to auditing mechanisms, the value of that “low barrier” is diminished.
On the other hand, the competitive advantage of professional content producers may actually be amplified. Compliance labeling combined with high-quality original content will create a new kind of trust premium. The future market landscape may bifurcate: on one side, low-cost, high-volume AI filler content; on the other, high-value, highly trusted authentic creations.
For enterprise brands, this means a more complex environment. Marketing content must simultaneously comply with AI usage standards (if AI tools were used) and convey authenticity value to consumers. Future advertising strategies will likely need to balance both dimensions—using AI for efficiency and relying on authenticity to build trust.
When AI production efficiency far exceeds what humans can imagine, “real vs. fake” is no longer a simple labeling problem—it is a systematic engineering challenge that requires multiple layers of technology working together. Digital watermarks address attribution. Spectral anti-counterfeiting addresses the unforgeability of physical laws. The audit chain addresses full-process traceability. None of these three dimensions can be missing.
And the starting point for all of this comes down to the most unassuming foundational components: a qualified narrowband filter, a precision-stable detector, and an accurately calibrated signal chain. The ultimate battle in technology is never in the code—it is in physics.
About Us
Founded in 2007, Hangzhou MULTI IR Technology Co., Ltd. is an optoelectronic technology enterprise integrating R&D, production, and sales. Its products are widely applied in aerospace, medical care, AR/VR, display imaging, photography, and other fields, steadily holding the position of the world’s largest spot supplier of optical components.
Media ContactCompany Name: HANGZHOU MULTI IR TECHNOLOGY CO., LTD.Contact Person: Media RelationsEmail: Send EmailCountry: ChinaWebsite: https://www.miroptech.com/
