On-device machine learning
Also written: On-device ML
AI that runs on your phone or computer, not on company servers.
On-device machine learning means that the artificial intelligence processing happens right on your personal device, like a smartphone or laptop. Instead of sending your data to a company's powerful computers in the cloud for analysis, the AI model works locally.
This approach is used to protect your privacy. For example, WhatsApp uses on-device AI to scan messages for scams without sending the message content to Meta's servers. This way, the AI can help keep you safe while your conversations remain private.
On-device machine learning involves running AI models, particularly for inference, directly on end-user hardware. This contrasts with cloud-based AI, where computations occur on remote servers. Key benefits include enhanced privacy, reduced latency, and offline functionality.
Tradeoffs involve limitations in model size and computational power compared to cloud resources, requiring optimized models (e.g., through quantization or pruning) and efficient hardware utilization. This approach is crucial for applications where data sensitivity or real-time interaction is paramount.
On-device AI is gaining prominence as companies like Meta seek to enhance user privacy and security in their applications. Running AI locally allows features like scam detection to operate without compromising sensitive user data, addressing growing privacy concerns.
- How We’re Building Scam Alert on WhatsApp With End-to-End Encryption and Verifiability GuaranteesMeta · Engineering at Meta ·
- Meta Meta AI
- Meta uses on-device AI in applications like WhatsApp to provide features such as Scam Alert, enhancing user safety while preserving message privacy.
- Cloud AI
- Cloud AI runs computations on remote servers, requiring an internet co
- Edge Computing
- Edge computing is a broader concept of processing data closer to
- On-premise AI
- On-premise AI refers to AI systems run within an organization's o
The concept of running computation locally on devices has existed for decades, but its application to machine learning models has accelerated with the rise of powerful mobile processors and the need for privacy-preserving AI.
On your smartphone, look for features that offer real-time assistance or privacy enhancements without an obvious internet connection. For example, some keyboard apps offer predictive text that learns your typing style locally, or camera apps might use on-device AI for scene recognition.
The Playbook: reusable how-tos, prompts, configs and tips pulled out of the news.
- PromptRole-play a student struggling with math for AI tutor simulationSimulate a student's math struggles and uncertainty for AI tutor evaluation. · 5 min
- TemplateReusable workflow templates for healthcare teamsTeams can create and share templates for common workflows like documentation or prior authorizations. · 15 min
- TemplateTemplate for building an AI tool tailored to your workA structured approach to defining and building a custom AI tool for specific business needs. · 15 min
- TemplateEvals CSV prompt list for agent skillsA CSV file template for listing prompts used in evaluating agent skills. · 20 min
- HeuristicDouble coding agent task completion with docs linksProviding coding agents with a documentation link significantly increases their task completion rate. · 2 min
Questions people ask
- Is on-device AI slower than cloud AI?
- It can be, as device hardware is less powerful than cloud servers. However, for many tasks, on-device AI is fast enough, and it avoids the network delay of sending data to the cloud.
- Does on-device AI mean no internet is needed?
- Not necessarily. While the AI processing itself happens locally, the app might still need an internet connection for other functions like downloading updates or syncing data.
First seen on Subvolts Aug 12, 2026. Updated Sep 15, 2026. Written by an AI model from the items above and reviewed against them weekly; report an error.