China Boosts Chip Design Protection; AI Model Faces Export Curbs
- tech360.tv
- 1 day ago
- 3 min read
China has initiated revised regulations to protect chip designs, a move intended to advance technological self sufficiency. Concurrently, Chinese artificial intelligence firm MiniMax has made its new H3 video model openly available but has imposed significant licensing conditions for users in major international markets. These actions highlight a period of evolving regulatory and commercial landscapes within the technology sector.
According to the South China Morning Post, Premier Li Qiang sanctioned new rules governing the protection of integrated circuit layout designs. These rules are scheduled to take effect in Oct. The revisions specify financial penalties for infringements and clarify rights pertaining to licensing, transfers, and co ownership. The objective is to expedite the commercialisation of new chip designs within the country.
The legal update arrives as Beijing increases its pursuit of technological self reliance. This aligns with the 15th five year plan, which seeks to refine and strengthen mature node processes, enhance capabilities in advanced process technologies, and accelerate the development of key equipment, materials, and components. And it aims to advance high performance processors and high density memory.
MiniMax, a Shanghai based organisation, made the weights of its H3 video model available to devs. The weights represent the core parameters encoding the model's intelligence, allowing devs to utilise, copy, modify, and distribute the software without direct payment to MiniMax. However, a licensing arrangement restricts free access in several key international territories.
Users discovered that the licence imposed geographical limitations, specifically affecting the United States, the European Union, the United Kingdom, and South Korea. To ensure compliance, devs operating in these regions must formally apply for a licence from MiniMax. The company indicated that authorisation would be granted following a review of deployment scenarios and confirmation of appropriate compliance controls and safeguards.
Access through the application programming interface, which involves a paid subscription via MiniMax's website or authorised third parties, remains available globally. MiniMax stated that the regional limitations stemmed from regulatory uncertainties and ongoing legal proceedings related to copyright. But in the United States, the restriction is directly linked to current generative video copyright litigation involving major Hollywood studios, as confirmed by Ryan Lee, MiniMax's head of developer relations.
The introduction of openly available video models has established a new arena for competition in generative artificial intelligence. Chinese devs are working to contend with United States rivals by providing models capable of local customisation and deployment. MiniMax's H3 model is presented as having advanced video generation features and broad hardware compatibility.
This model can generate videos up to 15 seconds in duration at 2K resolution, with a cost of 13 US cents per second. Benchmark platform Artificial Analysis ranked the H3 model first in video editing capabilities. It achieved a second place ranking in its ability to generate video from text, positioned behind Google's Gemini Omni Flash. The Google model costs 1 US cent per second but is limited to 10 second clips.
The imposition of licensing requirements occurs amidst increasing copyright disputes faced by generative artificial intelligence companies. Content owners are challenging the use of protected works in training data and the outputs generated by artificial intelligence. So, earlier, Walt Disney and NBCUniversal initiated legal action against Midjourney, a San Francisco based research laboratory. They alleged its artificial intelligence service produced unauthorised versions of copyrighted characters from franchises such as Star Wars and Marvel.
Chinese firm ByteDance previously faced copyright objections across Hollywood following the introduction of Seedance 2.0. Subsequently, the company rolled out the updated Seedance 2.5 model. This iteration claims the ability to create 30 second videos in a single run, though it is not made openly available.
China implemented new regulations to enhance chip design protection, aiming for technological self sufficiency.
MiniMax introduced its H3 video model with geographical licensing restrictions due to regulatory uncertainties and copyright litigation.
The H3 model provides advanced video generation but requires formal licences for devs in the United States, European Union, United Kingdom, and South Korea.
Generative artificial intelligence companies, including Midjourney and ByteDance, face increasing copyright disputes over content usage.
Source: SCMP
