Private AI for Side Projects: Keep Your Work Off Your Employer's Radar

import { Disclosure } from '@/components/Disclosure'

Bottom line up front: Every AI tool you use on a company device or with a work email is potentially visible to your employer — and could cloud your IP ownership. This guide shows you how to build a clean, separate AI workflow for side projects using local LLMs, a dedicated private identity, zero-knowledge storage, and research tools that don't cross-contaminate your work profile.

Last updated: 2026-06-30

Most tech workers understand the risk of committing side project code on a work machine. Fewer think about the subtler trails: the ChatGPT conversation where you sketched your SaaS idea, the GitHub Copilot autocomplete that ingested your unreleased architecture, the Google search history that maps your market research directly to your work account.

If you ever face a dispute over intellectual property, those trails matter. And even if you never do, giving your employer free visibility into your off-hours work is simply a privacy loss you don't have to accept.

The good news: building a fully separate AI stack for side projects is easier than it sounds, and it costs less than a gym membership. If you want the broader developer-focused version of this playbook, see our private AI toolkit for developers.

Why Your Current Setup Is Probably Leaking

Before the fix, understand what's exposed.

Company devices. Your employer may have MDM (mobile device management) software installed. Even without it, browser sync, Copilot for Business seat assignments, and shared SSO accounts all route through company infrastructure. Anything you access on that machine can be logged.

Work email. Signing up for AI tools with your work email makes you a company-associated user. If the company ever exports their ChatGPT Team or Claude for Business data, your side project conversations go with it.

Consumer AI tools on a work device. OpenAI, Anthropic, and Google all have clauses reserving the right to use conversation data to improve their models (unless you opt out on paid plans). That's a secondary issue — the primary one is that network traffic logs exist, and your IT team may or may not be monitoring them.

Cross-account search history. If you research your side project idea while signed into your work Google account, that search history is associated with you in Google's advertising systems and, in some cases, is surfaced in Google Workspace admin dashboards.

None of these scenarios require bad faith from your employer. It's just what happens when you don't separate identities. For a deeper look at the specific risks of using AI tools on a company-issued machine, see our guide on using AI tools safely on a work laptop.

Step 1: Build a Separate Identity Starting with Email

The cleanest way to separate your side project from your employer is to anchor it to a different email address — one not connected to any company account, not hosted by Google or Microsoft, and ideally encrypted end-to-end.

Proton Mail is the best choice here. Proton runs out of Switzerland under Swiss privacy law, uses zero-knowledge encryption (Proton can't read your emails), and you can create a free account that has no connection to your work identity whatsoever.

Use this email address for:

Signing up for AI tools you use on side projects
Domain registration and hosting accounts
Stripe, Lemon Squeezy, or any payment processor
Any SaaS product your project depends on

Affiliate Disclosure: This article may contain affiliate links. If you make a purchase through these links, we may earn a small commission at no extra cost to you. We only recommend products we genuinely believe in. This helps support our work and allows us to continue providing free content.

Once you have the Proton account, set up Proton Pass to keep all your side project credentials in a separate vault. Never let your browser or 1Password's work profile autofill across accounts.

The goal is that your side project's digital footprint looks like a different person — because from an account perspective, it is. Email separation matters more than most developers realize: see how your email gets embedded in every commit you push for why a single leaked commit can undo all of this.

Step 2: Run AI Locally for Code and Architecture Work

If you're using GitHub Copilot, Cursor, or any cloud-based coding assistant for your side project, you're sending your source code to a third-party server associated with your work identity (if you're using a company seat) or creating a personal trail attached to your side project's sensitive IP.

The alternative: run a local LLM on your personal machine.

Ollama is the fastest path. It runs open-weight models entirely on your hardware. Nothing leaves your machine. There's no account required, no API key, no usage logs.

For a 2026 MacBook with an M3/M4 chip or a mid-range PC with 16GB+ RAM, these models perform well for code assistance:

Qwen2.5-Coder:7b — strong on Python, TypeScript, and Go
DeepSeek-Coder-V2:16b — excellent on architecture explanations and refactoring (needs 12GB+ RAM)
Llama 3.1:8b — good all-rounder for brainstorming and planning

Install Ollama, pull a model, and connect it to VS Code via the Continue extension with your Ollama endpoint as the provider. You get autocomplete, chat, and inline edits — all local.

The privacy property is clean: your proprietary architecture, your unreleased feature ideas, and your business logic never leave your personal machine.

Step 3: Research Without Building a Tracking Profile

Here's the scenario you want to avoid: searching "SaaS pricing for project management for dental offices" while signed into your work Google account while also using the same browser profile you use for work. Now Google knows you're thinking about that market, and that signal is associated with your employer's account ecosystem. This isn't hypothetical — see how work searches leak company secrets to Google for how those signals get surfaced.

For side project research, you need a tool that either doesn't build a persistent profile or keeps it strictly separated.

Perplexity Pro is the most practical option for tech workers who need real-time information. Sign up with your new Proton Mail address — completely separate from any work identity. Perplexity's Pro tier gives you access to multiple model backends, real-time web search, and file analysis. On the privacy side, Perplexity's data handling is considerably leaner than Google's ad-driven model.

Use Perplexity (on your personal device, in your personal browser profile) for:

Market research and competitor analysis
Technical deep-dives on frameworks you're evaluating
SEO and keyword research for your side project's content
Pricing strategy research

For anything particularly sensitive — researching IP law, employment contract analysis, or anything you'd rather not have stored anywhere — pair Perplexity with your local Ollama setup. Download the relevant documents, run them through your local model, ask your questions entirely offline.

Step 4: Store Your Work in Zero-Knowledge Encrypted Storage

Where you keep your side project files matters as much as where you create them.

If your side project files live in your work Dropbox, Google Drive on your work account, or on a NAS that your employer set up — those files are not private. Even if your employer never looks at them, the access control isn't yours.

Tresorit solves this cleanly. It's zero-knowledge encrypted cloud storage, meaning even Tresorit's servers cannot read your files — only you hold the keys. It's built for exactly the kind of sensitive file handling a side project requires.

Set up a Tresorit account with your Proton Mail address. Use it to store:

Source code backups (though your primary repo should be on a personal GitHub/GitLab account, not a company one)
Business planning documents, financial models, and pitch decks
AI-generated assets you want to keep (local LLM outputs, Perplexity research digests)
Contracts with early customers or collaborators

The key property Tresorit provides isn't just encryption — it's provable separation. If an IP dispute ever arose, you'd be able to demonstrate that your work was stored in infrastructure you paid for personally, not co-mingled with employer systems.

Step 5: Protect Your Devices at the Network Level

Even with local LLMs and separate accounts, your home network router may still log DNS queries that reveal which services you're hitting. This is a lower-risk concern for most people, but if you work from home and your work machine and personal machine share a network, some MDM configurations can see traffic from adjacent devices.

The mitigation is simple: use Proton VPN on your personal devices while doing side project work. It encrypts your traffic at the device level and routes it through Proton's servers, preventing your ISP or a co-located work device from logging your connections.

You don't need a VPN for every task. Turn it on when you're doing sensitive research, making account registrations, or accessing your side project's admin dashboards from a shared network.

The Legal Layer: Review Your IP Assignment Clause

No amount of technical separation matters if your employment contract says your employer owns everything you create, regardless of what device you use.

Many tech employment contracts include "moonlighting" or IP assignment clauses that are broad. Some require you to disclose side projects to your employer, even if you have no intention of competing.

Before you put significant work into a side project, re-read your employment contract. Specifically look for:

IP assignment scope — does it cover only work-related inventions, or everything you create?
Disclosure obligations — are you required to report side businesses?
Conflicting interests clauses — does your project compete with your employer's market?

If the language is vague or broad, consult an employment attorney before launching. Technical privacy hygiene protects you from accidental data exposure — it doesn't rewrite contract terms. If you take on outside client work alongside your side project, the same gap applies to contract language: see what your NDA actually requires for encrypted storage.

The Full Stack, Summarized

Here's the complete setup in one place:

| What You Need | Tool | Why |

|---|---|---|

| Separate identity | Proton Mail | Swiss-law zero-knowledge email, no Google/Microsoft |

| Password separation | Proton Pass | Isolated vault for side project credentials |

| AI coding assistance | Ollama + Continue.dev | 100% local, no API, no usage logs |

| Private research | Perplexity Pro | Separate account, real-time web, leaner data model |

| Encrypted file storage | Tresorit | Zero-knowledge, provable ownership separation |

| Network privacy | Proton VPN | Encrypts device traffic on shared networks |

The total cost for the paid tier of these tools runs roughly $35-50/month. That's a small insurance premium against an IP dispute, and it's the same stack you'd want anyway if you're eventually launching something real.

What This Stack Doesn't Protect Against

To be clear about the limits:

Your employer's legal rights. If your contract says they own your side project, technical separation doesn't change that.
Your personal device's local storage. Files sitting in plaintext on your personal laptop aren't protected by any of the above unless you're also using full-disk encryption (FileVault on Mac, BitLocker on Windows).
Behavioral patterns. If you suddenly start shipping GitHub commits at 10pm every weeknight, no amount of tooling hides that from a suspicious employer.

Privacy tooling reduces your data footprint. It doesn't eliminate the reality that you're a person with patterns.

Start Here If You're New to This

If the full stack feels like too much to set up at once, do these three things first:

Create a Proton Mail account (free) and use it to register any new AI tools you use for side projects.
Install Ollama on your personal machine and pull qwen2.5-coder:7b. Use it for any coding you're doing on your project.
Move your sensitive documents out of work-connected cloud storage into Tresorit.

Those three steps eliminate the highest-probability exposure paths in under an hour.

Stay Ahead of What Your Tools Know About You

The landscape shifts constantly — new AI tools launch with different privacy postures, and employment law evolves. If you want a short weekly digest of what's changing in AI privacy (including new tools, policy changes, and threat patterns worth knowing about), subscribe below.

No spam. One email per week. Unsubscribe any time.