Japan eyes safety certification for AI tech

Seal of approval would limit user liability, promote development

DATE: 31/12/2016

Artificial intelligence technologies could undergo scrutiny to receive public certification as safe and secure, as Japan looks to create a system that limits user liability while encouraging a promising business field.

The system proposed by the Ministry of Internal Affairs and Communications would cover everything from AI-equipped computers like IBM's Watson to intelligent robots. No specific regulations or guidelines exist on research and development in the field.

The government seeks to address the risk of unrestrained AIs harming humans, or acquiring or leaking sensitive information.

The Japanese ministry, knowing that severe restrictions could hinder technological advancement, aims to design a certification program that allows safe products to stand out in the market.

AI technology receiving certification would need to be under constant supervision by a human or other AI, with a safeguard letting a human take control at any time. High security also is essential, to prevent AIs from being taken over or having information stolen in a cyberattack. Privacy measures, such as limiting access to or instantly anonymizing personal information, are on the checklist as well.



No comments: