Opera has initiated the early access rollout of WebAI within the Opera One browser. This feature enables websites and extensions to utilize AI models running locally on the user's device, facilitating tasks such as text generation and content classification.
The current preview is compatible with 64-bit Windows systems powered by Intel or AMD processors, as well as Mac computers equipped with Apple Silicon. Support for Windows on Arm is currently unavailable. Performance relies on the host machine's hardware capabilities rather than a dedicated NPU.
Supported models include Google's Gemma and Apple's local base model for Mac. Users must manually enable the feature, which triggers the download of models. Permissions allow users to restrict access to these models on a per-site basis.
Industry experts highlight potential security risks, noting that malicious sites or extensions could exploit local models or access sensitive data. Corporate IT departments may face challenges in managing and monitoring these local AI operations compared to standard network traffic.
While local processing may reduce reliance on cloud services for simple tasks like text summarization, it does not eliminate costs associated with hardware maintenance and system integration. Experts suggest a hybrid model, combining local AI for basic operations and cloud systems for complex tasks.
