Key Takeaways
OpenAI will reportedly open its third global office in Tokyo next month, boosting its presence in Asia as it expands internationally.
The choice reflects a growing emphasis on the Japanese market, with the AI developer planning to roll out Japanese language services later this year. It also comes at a time when the government is flirting with incorporating AI into its workflows.
During a press conference last year, Chief Cabinet Secretary Hirokazu Matsuno said the government was looking at how it could leverage foundation models and other AI tools to boost productivity.
However, he acknowledged that privacy and cybersecurity concerns would need to be resolved before government agencies could fully embrace the technology.
While government agencies could certainly benefit from integrating AI tools, they are also subject to enhanced security and compliance requirements that businesses and individuals might not be.
For example, the US House of Representatives recently banned staffers from using Microsoft Copilot on official computers.
In new guidance, the House Office of Cybersecurity deemed Copilot to be “a risk to users due to the threat of leaking House data to non-House approved cloud services.”
The decision reflects a significant challenge AI developers must overcome if they want governments to fully embrace chatbots: their poor track record of leaking sensitive information.
When it comes to OpenAI’s GPTs, for instance, researchers have found that attackers are often able to extract files and design prompts used to create the customized chatbots.
The opening of a new OpenAI in Tokyo comes as the Japanese AI market is expanding rapidly.
With plans to introduce comprehensive new rules for the sector later this year, the country is among the first movers in the global race to shape how AI is regulated.
Meanwhile, local startups like Airfriend have a headstart when it comes to Japanese language chatbots. But AI giants like OpenAI are preparing to muscle their way in.
Although there is already a Japanese version of ChatGPT, it is currently limited by a strong bias toward English in OpenAI’s training data.
Last year, CEO Sam Altman said the firm hopes to build “something great for Japanese people [and] make the models better for Japanese language and Japanese culture.” With the Tokyo office set to open soon, expect such enhanced-for-Japanese models to follow.