Market and Risk Mitigation for Generative AI – Part 2
We have previously discussed the market and risks associated with generative AI. This time, we will explain security measures against these risks.
Content generated by generative AI may contain inappropriate content such as racial discrimination or bias, concerns about leakage of company confidential information or personal information, and issues regarding the accuracy and reliability of the generated information. First and foremost, companies need to establish guidelines for governing generative AI and explain them to their staff. Subsequently, it is appropriate to take the following technical and operational approaches:
Risk Management for Generative AI
1. Prevention of Data Leakage through Monitoring and Restriction during Data Transmission
Introduce SSE (CASB) services such as Netskope and Zscaler to centrally manage the network of all servers and PCs, monitor transmitted data when confidential information is sent from the internal network to the external network, and block confidential information and personal information.
2. Prevention of Data Leakage through Monitoring and Restriction of Endpoint Inputs
Monitor information inputted on PCs via Endpoint solutions like Teramind to block the leakage of confidential information and personal information.
3. Usage Restriction through Access Control to Generative AI
Verify major generative AI services and adopt business-oriented generative AI services (which are often with fees) as internal standards, while blocking access to general services through various web filtering functions.
4. Creation of Generative AI Usage Guidelines and Training for Staff
In addition to traditional Cybersecurity Awareness Training, create guidelines for generative AI usage, conduct training on precautions when using generative AI services, and raise awareness among all users.
Step 1, the system known as SSE (CASB) is a cutting-edge security measure, and we strongly recommend it as an integrated system that not only protects against security risks specific to generative AI but also prevents other cybersecurity risks. Particularly, Netskope holds the top market share, and it is a highly reliable product that has undergone practical testing by the Japanese Ministry of Economy, Trade, and Industry.
Step 2, Monitoring from Endpoint involves monitoring user input as an additional function of systems that acquire operation logs for staff monitoring or forensic purposes and blocking them accordingly.
Steps 3 and 4 require regular maintenance and lecture work, which can be labor-intensive but are the most cost-effective and immediately implementable measures.
Various generative AI tools can streamline and accelerate business operations depending on their usage. However, due to the rapidly expanding market, security risks are also increasing. During the nascent stage of new services in a free market, vendors tend to prioritize functionality over safety to gain market share, leaving the security risks of each service to the users. We encourage you to consider security risk management for generative AI at this opportunity. If you have any concerns or questions, please feel free to contact us.
Continue Reading
OT Security Blog #4: Four Unseen Threats
The biggest threats to smart manufacturing are not the ones you can see. In the fourth installment of our OT Security Blog Series, we explain four OT security threats you can’t afford to overlook.
OT Security Blog #3: Five Alarming Facts
Today, smart manufacturing connects warehouses, production lines, and operational systems to the cloud—and exposes them to new cyber threats. In the third installment of our OT Security Blog Series, we explain five facts every industrial leader must understand to stay ahead.
OT Security Blog #2: AI and OT: Hype or Help?
AI is changing the cybersecurity game, but in OT, it’s not a magic fix. In the second installment of our OT Security Blog Series, we explore where AI is helping in OT—and where it still falls short.