Remini Enterprise Risk Assessment | AI Tool Risk

Back to AI Tools

Remini

Risk Overview

Moderate Risk

4.3

/10

Image Enhancement

Top Risks

Processes facial photos (biometric data) at scale; mandatory arbitration with class-action waiver for non-EU users; face data extraction even if limited to enhancement creates biometric risk profile; NowSecure mobile app risk rating flagged; aggregation of face data across millions of users creates high-value target.

Mitigations

Italian/EU company with direct GDPR applicability; explicit opt-in required for training; short defined retention periods (7-20 days); face data not used for identification; EU users exempt from arbitration clause; transparent policy on data sources.

Get started

Is your team using

Remini

safely?

Book a Demo Free Assessment

Highest Risk Categories

The highest-scoring risk categories identified in this assessment.

Legal Terms

For users outside EEA/UK/Switzerland: Terms require use of arbitration to resolve disputes rather than jury trials or class actions. Binding individual arbitration with class-action waivers. For EEA/UK/Switzerland users: GDPR-compliant terms without arbitration waiver—EU consumer protections apply. The arbitration/class action waiver is a major risk for non-EU users processing sensitive facial photos.

/10

Security Posture

No SOC 2 or ISO 27001 certifications publicly referenced. Processes sensitive facial biometric data at scale (millions of face photos). GDPR-compliant privacy framework in place. Bending Spoons has privacy team and legal infrastructure. No major public breaches found. Facial data explicitly stated to not be used for identification. Security practices appear adequate but no third-party certifications verified.

/10

Training Use

Remini explicitly states training datasets may include user-provided images where users have explicitly opted in to model training. Default is opt-in is NOT automatic—user must explicitly consent. Datasets come from internally developed, publicly available, synthetic, or third-party sources. Facial feature data (estimates of location and shape) is extracted but not used for identification/authentication. Opt-in for training is the positive design here.

/10