Chrome installed 4GB LLM on ~3B devices via update channel, using a generic ToS clause ("software sometimes updates automatically"). No consent prompt, no opt-in, no opt-out until Feb 2026. The update exemption in CASL covers bug fixes, not new AI model deployments.

4GB model with access to full browsing context — no scope limitation. Any website can invoke chrome.ai on the full DOM of the current tab. No restriction to "scam detection" use case. The stated purpose does not bound the data accessible to the API.

Opt-out default (must actively disable) until Feb 2026 — none before that. Privacy-protective default = opt-IN. Google inverted the default to maximize deployment speed. The fact that they added a toggle in Feb 2026 proves they knew a privacy-protective default existed; they chose not to implement it at launch.

No disclosure to data subjects at time of deployment. No update to privacy notices. No DPIA (Data Protection Impact Assessment) published. No record of processing activity updated. Google cannot demonstrate compliance — the accountability principle requires they can, on demand.

Chrome was consented to as a web browser. Gemini Nano deployment = new, materially different purpose (AI inference on browsing context). Prior consent for browser use cannot be inherited for LLM deployment. This is the "consent inheritance failure" theory — novel, no prior case has decided it.

chrome.ai API callable by ANY website, ANY extension — no enterprise access controls, no audit logging, no CASB visibility. The API surface was not in any enterprise threat model. EDR/XDR/SIEM tools have zero visibility into chrome.ai invocations. Google deployed before enterprise security tooling could adapt.

If chrome.ai processes payment page DOM (including Google Pay, merchant checkout, banking), Chrome becomes a PCI-scoped component. No merchant, bank, or payment processor has included Chrome + Gemini Nano in their PCI scope assessment. Every PCI-compliant organization using Chrome has an unassessed component in their CDE (cardholder data environment).

Canadian financial institutions using Google Workspace / Chrome on employee devices now have an undisclosed AI data processing arrangement. OSFI B-10 requires: due diligence, contract requirements, ongoing monitoring, regulatory notification for all outsourcing/third-party risk arrangements. None of these were applied to Chrome's Gemini Nano deployment.

If chrome.ai is exploited (malicious extension harvests data via chrome.ai API), the incident triggers mandatory breach notification obligations for: Google (as platform) · enterprise employers (for employee data) · regulated entities (within sector-specific timelines). No existing incident response playbook accounts for chrome.ai as an attack vector.