Use AI on confidential material, without handing it over

A good deck needs real context: revenue, client names, salaries, plans. Today, getting that deck from an AI means pasting all of it into a chat, where it leaves your control. Many company policies don't allow that, so people either bend the rule or do without the AI.

OpenGamma separates the design from the data. The AI gets placeholders and designs the deck. Your browser puts the real data back. You get a finished deck built on your real data, and the AI never saw it.

Your browser

Reads your document
Encrypts it with your key
Shows you what the AI will see
Fills the real values back in and writes the .pptx

encrypted
→

OpenGamma

Stores ciphertext it can't read
Stores the placeholder version you approved
Renders the deck design

placeholders
→

Your AI app

Sees [CLIENT_1] grew [PCT_1]
Writes the story and picks the layouts
Never sees the real values

Masked mode

The AI reads your text with names, companies, amounts, dates, percentages, emails, codes and your own terms swapped for tokens. Best when the wording matters.

Blind mode

The AI sees only structure: how long each block is, table shapes, column types and trends. Best for the most sensitive material.

Same output

Placeholders are sized like the real content, so the deck the AI designs fits once your values go back in.

Go deeper

Data storageExactly what we store and where, who could read it (including anyone with access to the server), backups, deletion, and what the design does not protect against.Read Encryption in detailA sentence followed step by step: block IDs, placeholder tokens, the ciphertext layout, how your passphrase and recovery key lock the vault key, and how export works.Read

Automated tests plant secret values in a vault and then search the database, file storage, logs and every response the AI receives. They run on every release.

Questions: shoumikgoswami@gmail.com · Terms · Privacy