LLM Capsule

A few weeks ago, I wrote about the AI paradox — everyone wants the productivity, but no one trusts where the data goes.

We built LLM Capsule to fix that. Local encapsulation, context-aware tokens, automatic restoration. The privacy problem? Solved.

But here's what we didn't expect: solving privacy didn't automatically make people use AI.

We deployed with real teams — public sector, insurance, industrial cybersecurity. The privacy layer worked. Compliance teams signed off. But actual daily usage? Way lower than we thought.

Turns out, making AI "safe to use" isn't the same as making AI "easy to use." And if it's not easy, people default back to doing things the old way — or worse, using personal accounts with zero protection.

Here's what we found, and what we changed.

1. "I just want to ask a question."

In our v1, you had to upload a document before you could start a conversation. That was the whole design — you upload sensitive files, we encapsulate them, you chat with the AI safely.

But we kept hearing the same thing: "I just want to ask the AI something first."

Sometimes people want to test a prompt before committing a file. Sometimes the question isn't about a document at all — it's about a process, a regulation, a quick lookup. Forcing a file upload before any interaction created a wall right at the front door.

We removed it. Now you can start a conversation with or without a file. Upload when you need to, not because the tool demands it.

Small change? Technically, yes. But it fundamentally shifted how people approached the tool — from "I pull this out when I have a sensitive document" to "this just runs in the background whenever I use AI."

That's the direction we're heading. LLM Capsule shouldn't be a separate destination. It should be a layer that sits in front of whatever AI you're already using — invisible until you need it, always protecting.

2. "I was working on something yesterday. I can't pick it up."

This one was more subtle. In v1, your past sessions were saved — you could see the full history. But you couldn't go back into a previous session and continue the conversation.

Think about what that means in practice. You're analyzing an insurance claim on Monday. You get pulled into something else. On Wednesday, you want to pick up where you left off — refine the analysis, ask a follow-up question, go deeper.

But you can't. You'd have to start a new session, re-upload the file, re-explain the context. So most people just... didn't come back.

If a tool can't support continuity, it can't become part of a workflow. It stays a one-off experiment.

We fixed this. You can now resume any past conversation exactly where you left it. Your context, your files, your thread — all intact.

This wasn't a feature request. Nobody said "I want session persistence." They said "I tried it once and didn't come back." We had to figure out why.

What we learned

The pattern is the same in both cases: privacy is necessary but not sufficient for AI adoption.

We solved the trust problem. But trust doesn't create a habit. People need to be able to reach for AI the moment a question pops into their head — no file required, no setup. And they need to come back the next day and pick up where they left off — like any real work tool.

If your product feels like a checkpoint — extra steps, extra screens, extra approvals — people will use it only when forced to. That's not adoption. That's compliance theater.

Real adoption happens when protection doesn't require changing behavior at all. You shouldn't have to switch to a different tool to be safe. The protection should come to you.

That's the direction for LLM Capsule. Not another AI interface competing with ChatGPT. A layer — a plugin — that sits in front of whatever AI tools your team already uses, so protection is automatic and invisible. Teams don't adopt a new tool. They keep using what they already use, and Capsule handles the rest.

What's next

Now that daily usage is growing, we're seeing new problems — the kind that only surface when teams use AI on real work, every day.

Every company defines "sensitive" differently. A hospital's secrets look nothing like a manufacturer's. Standard PII lists don't cut it — teams need to define their own rules in minutes, not months.

When AI processes real enterprise documents — tables, nested sections, cross-references — most approaches break the structure. The AI gets confused, the output is wrong, and nobody trusts it.

And teams don't want to be locked into one LLM. Different models are better at different tasks. They need one protection layer that works across all of them — GPT, Claude, Gemini, whatever comes next.

That's where we're taking this. LLM Capsule as a plugin layer that works across any AI tool, any model, any workflow. More in the next update.

We started with a frustrating observation.

Every team wants to use ChatGPT, Claude, or Gemini to move faster. But most companies are banning them — because the moment a sensitive prompt leaves your environment, it's gone. You can't take it back.

So what happens? Security teams block access. Employees use personal accounts anyway. And now you have a "shadow AI" problem that's even harder to control.

The real issue isn't AI. It's that the only options today are "block everything" or "risk everything."

We built LLM Capsule to break this deadlock.

The core idea is simple: what if sensitive data never had to leave your environment in the first place?

Most security tools just black out text — turn names into *** and call it a day. But that kills the context. The AI can't understand what you're asking, so the response is useless.

LLM Capsule takes a different approach. Before your prompt reaches any external LLM, it automatically detects sensitive information and replaces it with context-aware tokens — locally, on your side. The AI still understands the intent and generates a useful response. The original data is restored only after the response comes back to your environment.

This works across plain text, PDFs, reports, and structured documents with tables or mixed layouts. If there's sensitive data in it, Capsule catches it — without breaking the meaning.

What we're launching with

This is our v1, focused on the essentials that make adoption possible from day one:

• Real-time local encapsulation — sensitive data is detected and replaced before anything hits the LLM. No data leaves your environment.

• DIY policy setup — admins can define what counts as "sensitive" for their org in minutes. Custom keywords, regex patterns, category-level rules. No consultants needed.

• Full audit trail — every prompt, every encapsulation event, every policy applied. Your security team gets the paper trail they need to say yes.

Why now?

We're seeing massive demand from industries where data privacy isn't optional — it's law. Telecom, insurance, public sector, industrial cybersecurity. These organizations want to adopt AI yesterday, but they can't wait 6 months for a security review.

They don't need another tool that blocks AI. They need infrastructure that makes AI safe to use.

That's what we're building. Privacy shouldn't be a barrier — it should be the thing that finally lets your team go all-in on AI.