Opera has begun rolling out artificial intelligence models that run locally within its browser, allowing users to perform certain AI-powered tasks without sending the information being processed to external cloud servers.
The company has introduced support for WebAI, a proposed web standard, through an optional early preview in Opera One. The technology enables websites and browser extensions to access AI models running directly on a user's device for functions such as text generation and content classification.
The preview is available on 64-bit Windows computers powered by Intel or AMD processors, along with Macs using Apple silicon. Windows devices running Arm processors are not currently supported.
Unlike cloud-based AI services, which typically require data to be transmitted to remote infrastructure for processing, Opera's approach allows supported tasks to be handled locally. However, the company has not specified a requirement for dedicated neural processing units, and performance will depend on the hardware being used.
The feature is optional, with models downloaded only after users enable the preview. Opera has also introduced website-specific permissions to determine which websites can access the locally running models.
The browser currently supports models including Google's Gemma and Apple's on-device foundation model on compatible Mac computers. Opera plans to introduce additional models optimised for different devices as the technology develops.
The rollout could have implications for enterprises assessing AI adoption, particularly organisations handling confidential business information. Local processing may reduce the amount of sensitive information transmitted to external AI providers, although it does not eliminate security concerns.
Pareekh Jain, CEO of Pareekh Consulting, noted that local AI processing could reduce certain data exposure risks. However, monitoring AI activity across individual employee devices could become more difficult for enterprise IT teams.
Security concerns also extend to malicious websites and browser extensions that could attempt to manipulate local models or gain access to sensitive information.
Tulika Sheel, Senior Vice President at Kadence International, highlighted the importance of enterprise controls governing website permissions, model management and information access.
The development also raises questions about hardware requirements and AI-related expenditure. Smaller models running on existing computers could support tasks such as summarisation and rewriting without requiring immediate investment in dedicated AI PCs.
However, more demanding applications, including complex reasoning and advanced AI workloads, may continue to depend on specialised hardware or cloud infrastructure.
Local browser-based processing could also reduce some cloud AI usage and associated API costs. Nevertheless, organisations would still need to consider expenses related to model updates, device management, security and integration with existing systems.
Opera's WebAI preview remains an early-stage implementation, and broader enterprise adoption will depend on performance, compatibility and the ability to manage AI capabilities consistently across devices.
Disclaimer: This article may include information derived from interviews, press releases, public statements, research, company communications and other publicly available or third-party sources. Such material may be summarised, paraphrased or contextualised for journalistic and editorial purposes. All rights in third-party content remain with their respective owners.