We’ve improved our machine learning (ML) model for signup to deliver stronger protection against automated account creation while keeping friction low for legitimate users.
Note: This update applies only to the signup flow. There are no changes to the ML models used for bot detection in login or password reset flows.
Expanded detection signals:
The model now leverages user-agent–based signals, such as operating system and browser version data, to more accurately distinguish between human and automated signup attempts.
Smarter traffic classification:
An updated labeling strategy improves how the model differentiates between malicious and legitimate signup activity, helping it adapt more effectively to evolving attack patterns.
Optimized sensitivity settings:
Adjusted detection thresholds capture a broader range of bot activity while maintaining a low false positive rate, ensuring a smooth experience for valid users.
These enhancements strengthen the signup protection capabilities of Attack Protection, enabling more effective detection of automated signup attempts without adding unnecessary friction for real users.
The rollout is in progress for all Enterprise customers with the Attack Protection add-on and will complete over the coming weeks in line with individual release schedules.
For configuration guidance or to learn more about protecting your signup flows, please refer to our documentation or contact your account team.
Fetched April 18, 2026