Skip to main content
Elephas lets you ask questions across your client files while you decide where AI runs. Indexing happens on your computer. Always. For the most sensitive matters, Local AI writes answers on your computer, so the content of your questions and documents is not sent to any AI provider. Someone at your firm has probably pasted client material into a chatbot this month. Elephas gives you a better habit: answers from your documents, not the internet, with every claim cited back to your own files.

​
Before you start

Elephas is a tool, not legal or compliance advice. Your professional rules, client agreements and firm policies decide what you may do with client material. Elephas gives you the controls to follow them.

​
Steps

1

Create one Brain per client or matter

Use New Brain (⌘B on Mac, Ctrl+B on Windows) and name it after the client, case, engagement or deal. Ask only searches the Brains you select, so an answer about one client never draws on another client’s files. See Create and manage Brains.
2

Add the client's files

Add the PDFs, Word files, spreadsheets, scanned letters and meeting notes for that matter. Elephas reads them, runs OCR on scans and builds the index on your computer. Your files are not uploaded to be indexed. See Add files and folders.
3

Add client names to Protected Names

Open the Brain, click More, and choose Protected Names. Type each client name, counterparty, project code name or other term in Name or term, click Add, then Done. These are always replaced with placeholders before that Brain’s content is sent to cloud AI. See Protected Names.
4

Choose where AI runs

Open Settings › AI and, under How Elephas answers, pick the option that fits the material:The setting applies to all your Brains, and you can switch it at any time. See Choose how Elephas answers.
5

Keep personal information protection on

In Settings › AI, leave Protect personal information when using cloud AI on. It is on by default. See Personal information protection.
6

Check the shield and the badge on answers

Under each cloud answer, the shield (for example 3 details hidden) lists what was replaced before sending. The Answered by: badge shows which AI answered. See Review hidden details.
The steps are the same on Mac and Windows.

​
How it works

​
Custody of your files

Your files, the text Elephas extracts from them, the search index and your chats are stored on your computer, in an encrypted database with keys in Keychain (Mac) or Credential Manager (Windows). With local indexing, custody of your files never changes hands. See Where your data is stored. Personal Sync is off unless you turn it on. If you do, your Brains travel between your own computers end-to-end encrypted, and Elephas cannot read them. See Sync encryption.

​
What each AI option means for client work

In Local AI mode, your question and your documents are not sent to any AI provider. Elephas still checks your plan online about once an hour while you use it, and that check sends only a device ID. See Use Elephas offline.

​
What personal information protection does, and does not do

Elephas finds many common personal details on your computer, such as names, organisations, email addresses, phone numbers and account numbers, and replaces them with placeholders before cloud AI sees them. It does not catch everything. Protected Names closes the gap for the names you know matter. The topic of your question and the rest of each passage are still sent. For matters where any miss is unacceptable, use Local AI.

​
File names can identify a client

With cloud AI, the file names of documents in scope can be sent along with passages. A file called “Smith divorce settlement.pdf” says something on its own. Consider neutral file names, such as a matter number, for sensitive documents.

​
Team workspaces

In a Team workspace, the Team Owner sets the AI Policy for everyone: Disabled, Allowed, or Allowed with required redaction. With required redaction, personal information protection always runs on Team cloud AI requests, and if it can’t run, nothing is sent. Own API keys are not used in Team workspaces. See Team AI policy. The Team audit log records sign-ins, role changes, device approvals and Brain changes. Its entries name who and what, never what a document says. See Team audit log.

​
What Elephas does not promise

  • Elephas does not make your use of AI compliant with any specific rule, and holds no security or compliance certifications today.
  • Elephas makes no promise about professional privilege or confidentiality obligations. It gives you control over where your files are stored and where AI runs. Whether a given use fits your obligations is your judgement.
  • Cloud AI is not end-to-end encrypted. The gateway and the model provider read the passages they receive in order to answer.

​
Troubleshooting

Add it to that Brain’s Protected Names. From then on it is always replaced before the Brain’s content is sent to cloud AI. See Protected Names.
The local model has not finished downloading or setting up. Open Settings › AI and check its status. See Local AI troubleshooting.
Protection is on, but its models are not ready, so nothing was sent. Finish the download in Settings › Downloads. See First-run download failed.

​
FAQ

That is your call under your own rules and agreements. If no text from those files may reach an AI provider, use Local AI: indexing and answering both run on your computer. Elephas does not warrant privilege or confidentiality outcomes.
The AI option is a single setting in Settings › AI, not a per-Brain setting. Switch it before you ask about a sensitive matter, and check the Answered by: badge on the answer.
Elephas does not train models on your documents or questions. With cloud AI, the provider’s terms decide retention and training. See What data is sent to AI.
Not in your Personal workspace, which is yours alone. In a Team workspace, a Team Owner or Admin shares each Team Brain with specific members. See Team Brains.

What data is sent to AI

Use Elephas offline

Protected Names

Team AI policy