Jumio, headquartered in Sunnyvale, California and founded in 2010, uses artificial intelligence to compare a live selfie against the photo on an identity document in seconds. The system is designed to stop someone from leveraging a stolen ID or a fake image. By 2023 the firm reported processing more than 300 million verifications annually and had cumulatively confirmed over 1 billion identities.
The core technical step is facial matching. AI models, trained on a large corpus of identity documents and faces, extract a mathematical representation from the selfie and compare it to the photo on the ID. The system looks for structural features: the distance between the eyes, the shape of the jawline, and the contour of the nose. When the similarity score passes a threshold, the identity is considered matched.
The harder problem is liveness. A static photo or a video replay of someone else's face must not pass. Jumio's liveness detection earned iBeta Level 2 certification for passive liveness in 2021. Passive means the user never performs a challenge such as blinking, turning their head, or reading digits. Instead, the AI analyzes subtle cues in the image: skin texture, lighting reflections, depth information, and micro-movements that a photograph or a mask cannot reproduce. The system catches print attacks, video replay attacks, and 3D mask attacks without requiring the user to follow a single instruction.
Accuracy, speed, and edge cases
Jumio does not publish a single false acceptance rate or false rejection rate for all deployments. Accuracy depends on ID document quality, lighting conditions, and the specific configuration a client chooses. Still, the iBeta Level 2 certification signals that the passive liveness engine meets a recognized industry benchmark for spoof resistance.
Edge cases the system handles
The AI is trained to cope with poor lighting, shadows, and low-resolution selfie cameras. When an ID photograph is years old, the algorithms account for natural changes: weight gain, facial hair, wrinkles, and subtle shifts in skin texture. The system is not designed to distinguish identical twins. If two people share the same facial geometry, the biometric comparison yields a high match score for both, so identity confirmation relies on the document and any extra checks the client elects to run.
The end-to-end transaction usually completes within a few seconds. The user photographs an ID and then captures a selfie. The AI processes both images, runs the liveness check, and returns a pass or fail result. Speed varies with network latency and image quality, but the platform is built for real-time use at scale.
Commercial applications and regulatory compliance
Jumio's technology serves financial services, online gaming, sharing economy platforms, and healthcare. Banks and fintechs use the system for know-your-customer checks during account opening. Online gaming operators verify that players are of legal age and are who they claim to be. Ride-hailing and home-rental marketplaces use identity confirmation to build trust between strangers. Healthcare organizations verify patient identities for telehealth and prescription services.
Compliance certifications
Jumio holds ISO/IEC 27001:2013 certification for information security management, which covers the processes for protecting customer data. The business is also SOC 2 Type II compliant, meaning an independent auditor has verified that Jumio's security, availability, and confidentiality controls are designed effectively and operated over a period of time. The company does not disclose GDPR compliance as a separate certification, but SOC 2 Type II and ISO 27001 are routinely used to demonstrate data protection practices that align with GDPR requirements.
Centana Growth Partners acquired Jumio in 2016. The firm reported a net revenue retention rate of 123% in 2022, indicating that existing customers expanded their spending faster than the rate of churn.
Data retention and privacy for biometric data
Jumio's approach to biometric data is governed by its privacy policy and client contracts. The organization states plainly that it does not sell biometric data. When a user submits a selfie and an ID document, Jumio processes the images, extracts the facial template, and then deletes the images or retains them only as long as required by the client's policy or applicable law. The template itself is a mathematical representation, not a photograph, and is stored in encrypted form.
The retention period shifts by jurisdiction. In regions with strict biometric privacy laws, such as Illinois under the Biometric Information Privacy Act, contracts require clients to specify retention and destruction schedules. The ISO 27001 and SOC 2 certifications provide a framework for managing that data securely, but the ultimate obligation for compliance with biometric privacy laws rests with the client that collects the data.
Jumio does not use the biometric data for training its AI models without explicit consent. The training dataset was assembled from publicly available sources and from data supplied by clients under agreements that permit its use for model improvement. The privacy policy is updated periodically, and clients are advised to review the terms that apply to their specific deployment.
How the technology fits into the broader identity verification market
Jumio competes with providers such as Onfido, Mitek, and Veriff. The central differentiator is passive liveness certified to iBeta Level 2, which enables confirmation without requiring the user to perform any active step. That distinction matters for user experience: a person who never blinks, smiles, or turns their head is more likely to complete the flow without frustration.
The broader market is moving toward passive liveness as the standard, because active challenges can be fooled by deepfake videos that simulate the required motions. Jumio's approach, which analyzes micro-signals in a single image or a short video, is designed to resist those attacks. Processing more than 300 million transactions per year gives the business an expansive dataset for training its models, which in theory should improve accuracy over time.
The 2016 acquisition by Centana Growth Partners supplied capital to expand the product line and go-to-market efforts. The company has not disclosed whether it has been acquired again or has gone public since then. As of April 2025, it continues to operate as a private entity under the ownership structure that acquisition established.
Key facts
- Headquarters: Sunnyvale, California
- Founded: 2010
- Annual transactions (2023): Over 300 million
- Cumulative verifications (2023): Over 1 billion
- Liveness certification: iBeta Level 2 (passive liveness, certified 2021)
- Information security certification: ISO/IEC 27001:2013
- Audit compliance: SOC 2 Type II
- Net revenue retention (2022): 123%
- Acquisition: Acquired by Centana Growth Partners in 2016
- Sectors served: Financial services, online gaming, sharing economy, healthcare
Comparison of liveness detection approaches
| Method | User action required | Spoof resistance | Example provider |
|---|---|---|---|
| Active liveness (challenge-response) | Blink, turn head, read digits | Moderate; deepfakes can simulate responses | Onfido, Mitek |
| Passive liveness (no challenge) | None; system analyzes single image or short video | High; analyzes micro-texture, depth, reflections | Jumio (iBeta Level 2 certified) |
| 3D structured light / infrared | None; requires specialized camera hardware | Very high; measures depth map | Apple Face ID, some Android phones |
Frequently asked questions
Does Jumio store my selfie or ID photo?
Jumio processes the images to extract a facial template and then deletes the images or retains them only as long as required by the client's policy or applicable law. The template is stored in encrypted form. Jumio does not sell biometric data.
Can Jumio tell if I am using a photo of a photo?
Yes. Jumio's passive liveness detection, certified to iBeta Level 2 as of 2021, analyzes skin texture, lighting reflections, depth information, and micro-movements to detect print attacks, video replay attacks, and 3D mask attacks.
What happens if the lighting is bad or I have aged since my ID photo was taken?
Jumio's AI models are trained on a large dataset that includes poor lighting conditions and variations in appearance due to aging, facial hair, and weight changes. The system is designed to still produce a match if the underlying facial structure is consistent.
Does Jumio work for identical twins?
No. The biometric comparison relies on facial geometry, which is nearly identical for identical twins. In those cases, identity confirmation depends on the document itself and any additional checks the client chooses to run.
How fast is the verification?
The end-to-end process typically takes a few seconds. The user takes a photo of their ID and then a selfie. The AI processes both images and performs the liveness check in real time.
