Gemini Spark is your AI. Singular. Always.
Google I/O 2026 unveiled Spark. An always-on personal agent that reads your Gmail and Drive and acts on your behalf. It's the most ambitious agent yet from a hyperscaler. It's also a single-account product, with no shared family workspace and no PII redaction between your documents and Google's models. Archevi is the family-shared, privacy-isolated, Canadian-hosted alternative for the documents you wouldn't paste into a Google search bar.
May 19, 2026: Google I/O announced Spark. Single-user by design
Spark launched at Google I/O 2026 as a personal AI agent woven into Gmail, Drive, and Calendar. It can summarize your inbox, find documents you forgot you uploaded, and take actions across Google's surface. It's powerful, and tied entirely to your individual Google account.
Spark has no shared family workspace, no PII redaction layer, and inherits Google's standard data handling. Archevi gives every member of your household full AI on the same vault, with PII stripped before the model ever sees it, hosted in Canada.
What you actually pay, and what you pay with
Per user, per month for Spark (estimated based on Google AI Pro tier). Tied to individual Google account. Five family members = $100/month.
- Always-on agent across Gmail, Drive, Calendar. Your account
- No shared family workspace. Spark is account-scoped
- No PII redaction layer. Spark reads your raw mail and files
- Hosting in Google's US-based infrastructure
Per family, per month. One vault. Every member gets full AI on the same documents.
- Shared family vault. Upload once, every member can ask
- PII redacted before queries reach the language model
- Hosted in Canada (Toronto), PIPEDA-aligned
- Files never enter model training. No setting to turn off
What Archevi does that Gemini Spark doesn't
Family-shared vault
One upload, every household member can ask. Spark is built around a single Google account. There's no concept of "the family asks one question and gets one answer."
PII redacted before the model sees it
Names, SINs, addresses, account numbers are stripped before any query touches the language model, then restored in your answer. Spark reads your raw inbox and Drive. The model sees everything.
Canadian data residency (PIPEDA)
Your documents live in Toronto on PIPEDA-aligned infrastructure. Spark inherits Google's US-based hosting and standard data handling.
Built for the documents you wouldn't paste into Google
Wills, insurance policies, tax slips, medical records. Spark is brilliant at "find that flight confirmation". It's not the tool for "what does Mom's will say about the cottage."
Feature-by-feature comparison
| Feature | Archevi | Gemini Spark |
|---|---|---|
Shared family vault, one upload, all members can ask Spark is bound to one Google account | ||
PII redaction before queries reach the model Names, SINs, account numbers stripped automatically | ||
Canadian data residency (PIPEDA) DigitalOcean Toronto US-based Google infrastructure | ||
Files never used for model training Consumer-tier interactions may improve services unless opted out | ||
Always-on agentic actions across Gmail + Drive What Spark is built for | ||
Expiry tracking for passports, insurance, licences | ||
Per-document Q&A with cited passages | ||
Flat family pricing (no per-seat) $6 to $24/month for the whole household $20/user. 5 members = $100/month |
Why "always-on" needs guardrails for family documents
Spark is the most ambitious consumer AI agent ever shipped by a hyperscaler. Google I/O 2026 demoed it summarizing inboxes, finding contracts in Drive, and chaining actions across Calendar and Gmail. For day-to-day productivity inside one person's Google account, it's a meaningful upgrade.
The trouble is that "always-on across your Google data" is a different proposition for family documents than it is for the rest of your life. Spark, by design, runs against the account it's installed in. There's no concept of a family workspace where Mom uploads her will once and her three adult children can each ask questions about it from their own logins.
There's also no privacy isolation layer. When Spark answers a question about your tax slip, the language model sees the raw document. Names, SIN, account numbers, all of it. Google's policies cover what happens to that data, but the model itself sees everything. Archevi puts a redaction step between the document and the model, every time.
And Spark inherits Google's data-residency story, which is excellent for Google's purposes but worth a second look for Canadian families. Documents live on US-based infrastructure subject to US laws. PIPEDA gives Canadian families a clear answer. Archevi is hosted in Toronto.
Use Spark for what it's brilliant at: your personal inbox, your calendar, your Drive. Use Archevi for the documents your family shares. The ones you'd want every household member able to ask about, with PII stripped, hosted under Canadian law.
What Spark does better, and we'll say so
Spark is genuinely state-of-the-art. Google has model depth, ambient compute, and a head start on "agentic actions across your data" that nobody else can match in the consumer space. If you live in Gmail and Drive and want one AI to act across them, Spark is the answer, and it'll only get more capable.
Spark also benefits from the rest of the Google ecosystem in a way Archevi never will. Calendar awareness, Maps context, search history, Photos. All of those weave into Spark's reasoning. Archevi is purposely narrow: it knows your family's documents, not your wider life.
Archevi isn't a Spark competitor. It's the version you reach for when the file in question is a will, a tax slip, or an insurance policy. Files you wouldn't want "always-on" against, files your whole family shares, and files you want hosted under Canadian law with PII redacted at the model boundary.
Spark for your inbox. Archevi for your family's documents.
Start on the Core plan for $6 CAD/month. One shared vault, every member with full AI, PII stripped, hosted in Canada.