US Intelligence Agencies Issue Warning on China-Based AI Threat
The NSA, CISA, and FBI warned that China-based artificial intelligence firms are conducting industrial-scale knowledge distillation campaigns targeting leading companies in the US.
The US National Security Agency, CISA, and the Federal Bureau of Investigation have issued a joint cybersecurity advisory against knowledge distillation campaigns conducted by China-based artificial intelligence companies targeting leading AI firms in the country.
Details of the Joint Security Advisory
The National Security Agency, in collaboration with the Cybersecurity and Infrastructure Security Agency and the Federal Bureau of Investigation, has released a joint cybersecurity advisory.
This advisory aims to raise awareness against knowledge distillation activities targeting leading US artificial intelligence companies.
Misuse of the Knowledge Distillation Technique
Knowledge distillation is a machine learning technique that involves training a less capable artificial intelligence model on the outputs of a larger, more powerful model.
Although this method is a legitimate training process, it can be abused to acquire competitors' capabilities in a shorter time and at a lower cost.
Methods of China-Based Campaigns
China-based artificial intelligence companies are making efforts to systematically extract proprietary functions and capabilities belonging to US models.
These activities are carried out through various means such as APIs, cloud providers, and proxy networks to evade detection and bypass geographic restrictions.
Urgent Measures to Be Taken
Immediate actions for US artificial intelligence companies include implementing comprehensive detection and mitigation mechanisms.
Additionally, targeted intervention modifications must be deployed, and inter-agency intelligence sharing must be established.