📚 DA-CORPUS-BROWSER-LLM-RISK-001 · BROWSER LLM RISK CORPUS
CLASS ACTION PRECEDENTS · CASL ENFORCEMENT · GDPR DECISIONS · INSURANCE GAP ANALYSIS · NOVEL THEORIES
⚖️ ONBA CASE 🔩 DA-ENGINE 📚 DA-CORPUS 📋 CLO BRIEF ⚙ BONIXER ⛓ XML SPINE
1
CASL s.8 ENFORCEMENT ACTIONS · CANADA
Express consent required for computer program installation · Chrome = strongest CASL s.8 case ever available
CRTC v. Compu-Finder (2015)
First CASL enforcement action
$1.1M administrative penalty. Established: CRTC will pursue even without private class action parallel. Software installation without consent = CASL violation.
$1.1M penalty
CRTC v. Datablocks / Sunlight Media (2018)
Browser-based adware installation
Adware installed without consent via browser mechanism. $100K penalty. Confirms: browser-based installation is covered by CASL s.8. No browser exception exists.
$100K penalty · browser install = covered
CRTC v. Rogers Communications (2017)
WiFi injection without consent
Rogers injected code into users' HTTP traffic via WiFi without disclosure. Demonstrated: even legitimate companies face CASL exposure for undisclosed code execution in user's browser context.
Consent decree · compliance program
Chrome Gemini Nano → CASL s.47 Class
Ontario · OPEN · first-mover position
No prior s.8 case at this scale. Chrome install on ~6.3M Ontario devices. $200/device statutory damages. $1M/day ongoing. 3-year limitation period from first knowledge of harm (researcher report: May 2026).
~$1.26B gross Ontario exposure
CASL s.47 ONTARIO EXPOSURE CALCULATION
$1.26B
Gross statutory damages (Ontario devices alone)
Ontario Chrome users: ~3.5M · Devices per user: ~1.8 · Total installs: ~6.3M
$200 per installation × 6.3M = $1,260,000,000
Plus: $1M/day ongoing from date of first install (2024) + additional Chrome AI feature invocations
Practical settlement range: $50M–$300M (comparable to Sony $150M at lower scale)
2
PIPEDA CONSENT INVESTIGATIONS · OPC DECISIONS
Meaningful consent standard · Clearview directly on point · OPC has investigative jurisdiction
OPC v. Clearview AI (2021)
Closest Canadian precedent
Scraped billions of photos without consent. OPC: no valid consent basis; legitimate interest rejected. Federal Court upheld. Directly on point for Chrome: Gemini Nano processing browsing context = processing personal data without meaningful consent.
OPC finding + Federal Court uphold · DIRECTLY ON POINT
OPC v. Facebook / Instagram (2019)
Bundled consent insufficient
Instagram data use without specific consent. OPC: bundled consent insufficient; specific consent required for new data uses. Chrome parallel: generic ToS update clause ≠ consent to LLM deployment.
Bundled consent = invalid
OPC v. TikTok (2024)
Buried consent clauses insufficient
Data collection without adequate consent disclosure. OPC: buried consent clauses in ToS insufficient. Chrome's "software updates automatically" = buried, generic, inadequate.
ToS burial = no consent
OPC Consent Guidelines (2018)
5 principles Chrome violates
1. Distinct purpose per consent ❌
2. Plain language ❌
3. Sensitive data = opt-in ❌
4. No bundling with unrelated service ❌
5. Easy withdrawal ❌ (until Feb 2026)
Chrome fails all 5 OPC principles
3
US CLASS ACTION PRECEDENTS · SONY LINEAGE
CFAA theories · trespass to chattels · consumer protection · Chrome is stronger than all predecessors
Sony BMG v. Class (2005–2007)
The controlling precedent
XCP rootkit via music CDs. CFAA + trespass to chattels + consumer protection. $150M settlement (cash + downloads + recall). FTC + state AG actions. Chrome is stronger: API surface (Sony had none), larger scale, persistence behaviour more aggressive.
$150M settlement · FTC consent decree
Lenovo Superfish Class (2015)
Shopping feature ≠ justification
HTTPS man-in-the-middle via adware. $3.5M + FTC consent order. Key: Lenovo's "shopping feature" justification ≡ Google's "scam detection" justification. Lenovo lost. The justification doesn't cure the consent failure.
$3.5M + 20yr FTC monitoring
In re Facebook Internet Tracking (9th Cir. 2022)
Re-download after deletion = violation
Key ruling: users have reasonable expectation that software they've rejected won't continue to execute code on their behalf. Chrome re-download after deletion = continuing to execute code after user exercised choice to remove. 9th Circuit ruling applies directly.
Re-download = ongoing violation
In re Google Street View (2013)
Google "collect first, defend later" pattern
Street View captured payload WiFi data. $7M settlement with 38 state AGs; private class certified. Establishes: courts repeatedly reject Google's "we didn't intend harm" defence. Pattern evidence for Chrome case.
$7M · 38 state AGs
4
GDPR ENFORCEMENT · EU CLASS ACTION SCAFFOLD
Meta fined €390M for bundled consent · Chrome uses same model · $14B theoretical maximum exposure
Meta GDPR Consent (Irish DPC, 2023)
€390M for bundled consent
"Take it or leave it" consent = not valid consent under GDPR. "Use Chrome = accept Gemini Nano" is exactly the bundled consent model fined €390M. Chrome parallel is direct.
€390M fine · bundled consent invalid
Google Analytics Decisions (2022-2023)
Austrian, French, Italian, Finnish DPAs
All found Google Analytics data transfers unlawful. Establishes EU DPAs will act against Google data practices. Chrome Gemini Nano + Article 25 failure = stronger case than Google Analytics.
Multiple DPA enforcement actions
GDPR Maximum Exposure:
Alphabet 2025 revenue: ~$350B · GDPR max: 4% global turnover = ~$14 billion theoretical maximum
Practical expectation: DPA negotiated settlement + binding corrective measures + mandatory Privacy by Design implementation
EU staged rollout (DMA compliance suspected) is evidence Google knows the deployment is legally problematic
5
CYBER INSURANCE GAP ANALYSIS · THE RAINCHEQUE
Unmodelled systemic risk · correlated exposure · no reinsurance tower prices this
COVERAGE GAP — chrome.ai API abuse:
"Data breach" = unauthorized access ✓ — but not by the software vendor
"Network security failure" = security control failure ✓ — but chrome.ai abuse isn't a failure, it's working as designed
"Privacy liability" = failure to protect ✓ — but Google is the vendor, not the enterprise

RESULT: No standard policy language cleanly covers chrome.ai exploitation as attack vector.
Enterprise's cyber insurer will dispute coverage. Google will argue it's not their liability. Gap in the middle.

REINSURANCE TOWER: Primary ($0-25M) → First excess ($25-100M) → Upper ($100M+) → Cat cover ($1B+)
Correlated chrome.ai exploit at enterprise scale exhausts all layers simultaneously. No model predicts this.
NotPetya comparison: $10B losses, excluded as war. Chrome: commercial software, no war exclusion. Reinsurers are exposed.
6
NOVEL LEGAL THEORIES TO DEVELOP
Not in any prior case · develop these now to own the precedent
THEORY 6.1
"Living Payload" Continuous Liability
Prior law treats software installation as point-in-time. chrome.ai is a living payload: ongoing API invocations, ongoing data processing, ongoing Google model updates — all without additional consent. Each invocation = new violation. CASL: $200/violation × daily invocations across all Chrome users = potentially astronomical continuous liability.
THEORY 6.2
"API Surface Duty of Care"
When a software vendor installs a model with an external API on user devices without consent, they assume a duty of care for that API's security. Google deployed the model, designed the API, failed to notify enterprise admins, failed to provide enterprise control mechanisms at launch. If API is exploited: duty → breach → damages.
THEORY 6.3
"Consent Inheritance" Failure
Google claims 2024 ToS click-through covers 2026 LLM deployment. Legal challenge: consent must be contemporaneous and specific to purpose. A 2024 generic update clause doesn't cover a 2026 AI model deployment. OPC 2018 guidelines + GDPR Article 7(3) expressly support this. Prior consent cannot be inherited for new, materially different uses.
THEORY 6.4
"Browser as Regulated Infrastructure"
Forward-looking: if browsers host AI models with enterprise API surfaces, they cross from "software" to "infrastructure" subject to financial/healthcare/critical infrastructure regulation. Develop now to support regulatory advocacy. Position EOSE's SBIL (Sovereign Browser Intelligence Layer) as the compliant enterprise-managed alternative.