China's Second AI Patent Infringement Case: Minimax and Hailuo AI in the Spotlight

A month ago, I just wrote about Zhipu AI’s first patent infringement case. Didn’t expect the second case to come so quickly—this time it’s Minimax and Hailuo AI.

Honestly, this news surprised me a bit, but thinking carefully, it was expected. Given where China’s AI large model industry has developed, patent wars were almost inevitable.

Case Background

According to public information, this lawsuit’s core involves patents related to “model training methods.” The plaintiffs allege that Minimax and Hailuo AI used patent-infringing technical solutions in their product development.

Specific technical details haven’t been disclosed yet, but based on industry practice, such cases typically involve:

• Training data processing methods: like how to clean, annotate, and augment data
• Model architecture design: like attention mechanism implementations
• Inference optimization techniques: like specific quantization, pruning, and distillation implementations

These patents’ characteristic: technical details are often hidden deep in code, hard for outsiders to discover. Once products are formed and commercialized, infringement becomes hard to “hide.”

Why Now?

This timing is interesting.

On one hand, the large model industry has passed the “wild growth” stage. In 2023-2024, everyone was racing against time and market, few paid attention to intellectual property issues. Now the market landscape is basically formed, funding secured, teams assembled—patent issues naturally surface.

On the other hand, large model companies’ valuations are generally high. Zhipu AI went public with a market cap over 50 billion HKD, Minimax’s valuation is also in the billions USD. Such high valuations mean intellectual property value is amplified—patent litigation is no longer just “technical disputes” but real money interest battles.

Another possible reason: investor pressure. VCs increasingly focus on IP issues, especially in technology-intensive industries. If companies have IP risks, investors demand quick resolution. This could be an important factor driving litigation.

Industry Impact

This case might become a “watershed” for China’s AI large model industry.

First, it sends a clear signal: large models aren’t “lawless zones.” In the past few years, many thought the AI industry evolved too fast for patent protection to keep up. But now it seems the legal system is catching up quickly. In the future, every large model company needs to take IP issues seriously.

Second, this will accelerate the industry’s “patent arms race.” Like the smartphone industry, large model companies will start frantically filing patents and building patent pools. Partly for defense, partly to accumulate bargaining chips for future cross-licensing and litigation.

There’s another overlooked impact: legal risks for open source projects. Many domestic large models are developed based on open source projects. If open source projects themselves have patent issues, downstream companies could get “caught in the crossfire.” This issue isn’t discussed much in open source communities, but legally it’s a real risk.

Corporate Response Strategies

If you’re an AI startup, what should you do now?

First, IP due diligence. During technology selection, assess whether IP risks exist. Especially for teams developing based on open source projects, carefully review project licenses and patent declarations.

Second, build a patent defense system. Not to sue others, but to have enough “ammunition” to protect yourself in litigation. Specifically:

• Actively file patents for core technologies
• Monitor competitors’ patents
• Establish internal patent compliance processes

Third, technology diversification. Don’t put all eggs in one basket. If a technology has IP risks, have backup plans. This is especially important in AI—most problems have multiple technical approaches, no need to stick to one path.

Deeper Issues

Behind this case lies a deeper question: should AI patents be protected?

Some argue AI technology develops too fast; patent protection hinders innovation. Patents filed 20 years ago might be completely obsolete now. Using such patents to sue equals “harvesting” later developers’ achievements.

Others argue without patent protection, no one will invest huge sums in R&D. AI industry burns money so fast, without IP protection, investors’ money might go down the drain.

Both views have merit. My take: the patent system needs to evolve, especially for rapidly iterating industries like AI. For example, shortening AI patent protection periods, or raising “non-obviousness” examination standards.

Of course, this requires joint effort from legal and industry communities. Short-term, we’ll likely see more patent litigation cases. This is both a sign of industry maturity and growing pains.

Final Thought

Regardless of this case’s outcome, it will be an important milestone in Chinese AI industry history.

For practitioners, it’s time to take intellectual property seriously. Don’t wait for a court summons to remember the word “patent.”

I’ll continue following this case’s progress. If new information surfaces, I’ll write another detailed article.