Russia to test AI models for threats to national security – media
The Russian government plans to check artificial intelligence models to ensure they do not pose “threats to national security and defense,” business daily Kommersant has reported, citing a new federal project on digital public administration.
The government earlier said it was seeking to accelerate the introduction of AI technology in areas including public administration and defense.
The planned security checks will apply to AI models that detect patterns based on government data sets, according to Kommersant. During the process of AI model training, data is fed into algorithms to help the system refine itself, and produce accurate responses to queries.
The government will allocate over eight billion rubles (around $72 million) by 2030 to develop software that will analyze AI models for safety, the paper reported. The Federal Security Service (FSB) will be responsible for the project, and the first such software system is expected to be introduced in 2027-2028. By 2030, a total of five AI models will reportedly be approved for “safe operation.”
Russia’s Ministry of Digital Development did not specify what kind of government data will be used in the AI model training. In June, the ministry approved the procedure for disclosing such data to companies that train AI models to improve service provision. The outlet reported last month that leading network providers and banks would gain access to data such as the property registry, passport expiry dates, data from the state pension system, and vehicle registry information.
Currently, AI models “obtain safety validations by testing and applying various scenarios to identify weaknesses,” Timofei Voronin, digital data expert from the Moscow State University, told Kommersant.
The government plans to employ AI models to facilitate decision-making and verification in areas such as the economy, public administration and defense. Last week Prime Minister Mikhail Mishustin announced the allocation of the equivalent of $172.5 million for domestic production of microchips.
