Cryptocurrency exchange Coinbase has patented a self-learning compliance enforcer that shutters “bad” user accounts.
The U.S Patent and Trademark Office-issued patent, published on Nov. 19, describes an automated system and accompanying scoring mechanism that together root out non-compliant user accounts – specifically ones suspected of trafficking in illegal activity.
It assigns an overall compliance score from a variety of factors, according to the patent description. Some are user-inputted or executed data points, including user age, account balance, transaction volume, location, verification history and the number of devices with access.
Other factors are calculated by the exchange itself. The compliance enforcer also measures “the level of due diligence that has been performed on the respective account,” and any past history of compliance review in determining “whether the account is bad or good.”
Good accounts pass through untouched. Bad ones are suspended and referred to law enforcement authorities if the transaction involves more than $2,000, the patent indicates.
Investigators have an override function allowing them to bypass the suspensions, according to the patent
The system learns to differentiate from a training set. It constantly updates its compliance model from the data collected and the accounts flagged.
While it is unclear whether Coinbase intends to ever implement such a system, the exchange has a strict set of know-your-customer and anti-money laundering policies meant to help it abide by U.S. and international laws, which this system could be designed to aid in.