Rakuten's AI Model Controversy: Innovation or Imitation?

Japan's tech community is embroiled in debate after Rakuten Group unveiled what it billed as the country's most advanced homegrown AI system. The 70-billion-parameter model, developed under METI's GENIAC project, promised to showcase Japanese technological prowess. But within days, eagle-eyed open-source developers raised red flags about its origins.

The Smoking Gun

Buried in the model's architecture lay undeniable evidence: configuration files still referenced "DeepseekV3ForCausalLM" - a model created by a Chinese development team. Rather than building from scratch as claimed, Rakuten appeared to have simply retrained this existing framework with Japanese data.

"It's like repainting a Ferrari and calling it your own design," commented one GitHub user analyzing the code. The discovery sparked heated discussions about what truly constitutes "self-developed" technology in today's interconnected AI landscape.

Transparency Troubles

The controversy deepened when examining Rakuten's communications:

• Selective Disclosure: Press materials vaguely referenced "open-source community integration" without acknowledging the Chinese model's foundational role
• License Limbo: Initial releases allegedly omitted required MIT license documentation, only adding compliance notices after community outcry

Legal experts note that while fine-tuning existing models isn't prohibited per se, ethical norms demand clear attribution. "The issue isn't building on others' work - that's how science progresses," explains Tokyo-based IP attorney Kenji Watanabe. "It's about whether companies are transparent when they do so."

Industry Implications

The incident highlights growing tensions between:

• Nationalistic tech ambitions
• The inherently collaborative nature of AI research
• Corporate pressures to demonstrate proprietary breakthroughs

As governments worldwide pour billions into domestic AI capabilities, this case may prompt tougher standards for what qualifies as "national" technology. Meanwhile, Rakuten has yet to address why Deepseek references remained or explain the initial license omissions.

Key Points:

• Rakuten's "Japanese" AI model shows striking similarities to China's DeepseekV3
• Configuration files and architecture suggest direct adaptation rather than original development
• Disclosure practices and license handling raise transparency concerns
• Incident reflects broader tensions in global AI development ecosystems