When you open the LINE app, many of you may have noticed an unfamiliar icon. This is the entrance to LINE Yahoo's long-awaited new AI brand ``Agent i'' on April 20th.
Agent i is a familiar AI agent that can be summoned with a single tap from LINE and Yahoo! JAPAN, which we use every day. The biggest feature is that it can be operated intuitively without inputting instructions (prompts). For example, with the recipe function, you can simply take a picture of the contents of your refrigerator with your smartphone and send it to the app, and the AI will recognize the ingredients and suggest a menu that can be made within 15 minutes using what you already have.
This is a convenient feature that frees you from the hassle of making daily menus. However, some people may stop when faced with this revolutionary convenience. "Is it really safe to send photos of my living space or personal concerns to AI?"
→A mysterious icon “Agent i” appears on LINE This is actually “your personal secretary” How will it enrich your daily life?
Past personal information leak issues that are still fresh in our minds
It's no wonder users have concerns about privacy. This is because LINE Yahoo is facing a large-scale personal information leak problem in the fall of 2023.
At the time, a company outsourced to South Korea's NAVER Cloud, an affiliated company of the company, was infected with malware. From there, it was accessed illegally through the network, resulting in the leak of personal data of approximately 520,000 users, including their service usage history.
The Ministry of Internal Affairs and Communications and the Personal Information Protection Commission took this situation seriously and issued strict administrative guidance and recommendations. LINE Yahoo is currently in the process of taking fundamental measures to prevent a recurrence, such as separating the network from NAVER, which was a weak point in the system structure at the time, and terminating the outsourcing of system management.
→Ministry of Internal Affairs and Communications requests LINE Yahoo to provide “administrative guidance”, protect “secrecy of communications” and ensure “cyber security”
→ Approximately 300,000 personal information leaked from LINE Yahoo; PC infected with malware → Unauthorized access caused
→LINE Yahoo apologizes for displaying incorrect photos, saying ``developer's understanding was insufficient'' ``Trust has been seriously damaged,'' Ministry of Internal Affairs and Communications says
Against this background, the new Agent i, which exchanges extremely private information such as the contents of the refrigerator, has been introduced. It is no surprise that there are voices of alarm.
Where will the data sent by the new AI be managed?
So where and how is the data that we upload to Agent i, such as the photos we have uploaded and the consultation history we have exchanged, managed? I asked this question directly to LINE Yahoo! Public Relations.
First, the most important thing to consider is where the data is stored. According to LINE Yahoo! PR, data such as usage history and generated responses on Agent i is "mainly stored on servers in Japan."
However, not everything is completed domestically. ``Some data is also stored at our U.S. subsidiary after taking strict safety management measures based on our privacy policy.'' Although some parts of the system still use overseas servers, at least it clearly states that the system will be operated from a U.S. base under strict control, rather than the South Korean server that was the site of past leaks.
Public relations response: ``It will not be used as training data for AI.''
Another concern many people have when using generative AI is whether their data will be used for the AI's learning and output as answers for someone else.
LINE Yahoo! PR clearly denied this as well. Behind the scenes, Agent i uses multiple world-class large-scale language model (LLM) technologies, including those from OpenAI and Google, depending on its functionality. However, it asserts that ``the data entered by users will not be used for learning purposes by the LLM vendor.''
In other words, there is no need to worry that the photos of the refrigerator we send for tonight's menu or the chat history of personal concerns will be used as training data to make the AI smarter without our knowledge.
Balancing convenience and privacy is important
In the future, Agent i will continue to add new features, such as being able to perform complex tasks on your behalf and read the atmosphere in LINE chat rooms.
On the other hand, in response to concerns about how privacy will be ensured when using the service, LINE Yahoo's recent stance by clearly delineating data storage within Japan (partially in the US) and non-use for learning data should be a major source of reassurance for users.
