The South Korean government and the U.S.-based multinational company Meta have agreed to use facial recognition technology to tackle the problem of deepfake ads on Meta’s social media platforms.

The Personal Information Protection Commission announced on Thursday that it plans to work with Meta, which operates Instagram and Facebook, to ensure the prompt removal of any advertisements that contain deepfake images.

Under the agreement, Meta must immediately delete any information about facial features that it extracts in this process and must only check such information for celebrities.

Also, Meta must not use facial recognition technology for any purpose other than confirming the identity of celebrities whose images appear in ads or on social media accounts and must submit its server log records to the Personal Information Protection Commission.

Meta also agreed to inform users in its privacy policies that profile photos and other information they make public can be used in this way to detect phishing ads or accounts.