Where the data goes
Nowhere. Document content, inference, and storage all stay on the device. There is no upload step to audit and no retention policy to negotiate, because Foxit never receives the document in the first place.
Some files can't go to a cloud service. Client records, patient files, sealed filings, anything under a confidentiality clause or a data residency rule. Local AI runs on your own hardware through Microsoft Foundry Local, so inference, document processing, and storage all stay on the machine. Nothing crosses a network, which means nothing crosses a border.
Local AI is included at no extra cost in Foxit PDF Editor and the free Foxit PDF Reader on desktop.
Microsoft Foundry checks your hardware and shows only the models your machine can actually handle. Better hardware unlocks larger models and better output quality.
Pick one from the list and download it through Microsoft Foundry Local. Download several and switch between them whenever you like.
One setting in AI Assistant. From that point the work happens on your device, and you can switch back to Cloud mode any time a task needs the full feature set.
An internet connection is needed for the initial model download only. Everything after that works offline.
Local AI runs on Microsoft Foundry Local, the same on-device model runtime Microsoft built for AI PCs. Foxit is a Microsoft Solutions Partner for Azure and a Microsoft AI Cloud Partner, and the model catalog, the device capability check, and the download pipeline all come straight from Microsoft rather than from a Foxit-built substitute.
That shortens a security review. The models come from Microsoft's catalog, the runtime is Microsoft's, and there is no Foxit inference service anywhere in the path. There is no third-party AI vendor to assess, because no third party ever receives the document.

Azure
PDF Editor

AI Cloud
Runs on AI PCs from Foxit OEM partners including Dell and Lenovo.

Local models are smaller than cloud models, and some capabilities need more than text. Here's the honest split, so nothing surprises you after the security review is done.
Unavailable capabilities are hidden or disabled in the interface rather than failing when you try them. Switching back to Cloud mode restores everything.
Rather than a downloaded local model, point Local AI at an OpenAI-compatible model you already run or pay for. Device performance stops being the limit, and document processing and storage still stay on your machine. Declare whether your model supports tools and vision during setup, and Foxit enables the matching capabilities.
Requests go to the external service you configure. Foxit tells you that plainly during setup.

Financial services, healthcare, legal and compliance, defense and government, and any team handling documents covered by a confidentiality obligation or a data residency requirement. When a document can't be uploaded, it has been getting no AI at all. Local AI is the alternative to that, without a new vendor in the stack.
Twelve desktop PDF editors compared. Foxit is the only one that runs AI entirely on your device.
Comparison includes Adobe Acrobat Pro, Nitro, PDFelement, UPDF, LightPDF, PDFgear, Tungsten Power PDF, PDF-XChange Editor, PDF Studio, pdfFiller, and Systweak. Accurate as of July 2026.
Intel-based Macs are not supported. The web-based Foxit Cloud Editor does not support Local AI.
Local AI was built so the answers are short.
Nowhere. Document content, inference, and storage all stay on the device. There is no upload step to audit and no retention policy to negotiate, because Foxit never receives the document in the first place.
Local AI runs inside the Foxit PDF Editor or Foxit PDF Reader your organization has already evaluated. No new vendor, no new contract, and no new data processing agreement for the AI itself.
On any device that meets the requirements, including machines in air-gapped or network-restricted environments. Once a model is downloaded, Local AI needs no connection at all.
Included with every Foxit account
FREE
Foxit Local AI is a mode within Foxit PDF AI Assistant that runs AI models directly on the user's device instead of in the cloud. Models are downloaded through Microsoft Foundry Local, and inference, document processing, and data storage all take place on the machine. No document content is sent to Foxit or to any external service.
No. Foxit Local AI is included at no additional cost in Foxit PDF Editor and in the free Foxit PDF Reader on Windows and macOS desktop, and it consumes no AI credits. Because processing happens on the user's own hardware rather than on a cloud service, there is no per-use charge.
Foxit Local AI runs on Windows desktops and on Macs with Apple silicon. Intel-based Macs are not supported. Microsoft Foundry checks the device and displays only the models it can run, so higher-performance hardware unlocks larger models and better output. Devices unable to run any supported model can use Bring Your Own Model instead.
Local models handle text-based work including questions, summarization, text generation, AI Bookmark, and document translation. Capabilities requiring cloud services or multimodal models are unavailable in Local mode, including Read Aloud, Smart Podcast, image and chart generation, editing a PDF by chat, formula recognition, web search, MCP connectors, and Skills. Switching back to Cloud mode restores them.
No. Foxit Local AI performs all inference, document processing, and storage on the user's own device. Document content is never transmitted to Foxit or to any external AI service, and no internet connection is required once a model has been downloaded through Microsoft Foundry Local. Bring Your Own Model is the one exception: requests go to whichever external service the user configures, which Foxit states during setup.
Local AI is included in Foxit PDF Editor and the free Foxit PDF Reader on desktop. Download a model, switch to Local mode, and disconnect from the internet if you want to prove it.