The Japanese Rakuten Group's recently released AI model
Rakuten Group claimed to have launched "Japan's largest" AI model, but it was quickly discovered by tech enthusiasts that the model was actually trained based on the DeepSeek-V3 model from DeepSeek. While fine-tuning an open-source model is a common practice in the industry, Rakuten's approach led to multiple controversies:
Compliance Gaps: In the initial release version, Rakuten removed the original MIT Open Source License file (LICENSE). According to the agreement, any derivative works based on this code must retain the original copyright notice and license statement.
Emotional Backlash: Netizens in the Japanese region pointed out that Rakuten used substantial computing resources supported by the Japanese government but failed to present fully self-developed results, a practice that was emotionally unacceptable to local residents.
After being caught by netizens and sparking widespread discussion, Rakuten quickly took action:
Re-Declaration: Rakuten published a file named NOTICE in the open-source repository, which fully included the original statement of DeepSeek-V3.
Compliance Status: Since the MIT license does not strictly require the specific naming of the declaration file, Rakuten's remedial measures have returned to compliance on a legal level, but it is still criticized by the community as "lacking transparency."
Although the related technical competition did not prohibit fine-tuning based on open-source models, Rakuten's "hiding" behavior was evaluated by the community as "lacking sincerity." This incident has once again triggered deep discussions about how companies should balance self-research and development with the use of open-source achievements when enjoying government subsidies.
Currently, Rakuten Group has not provided further explanations for the specific reasons behind removing the license.
