Skip to content
RiskTemplates · The Daily Brief Saturday, November 21, 2026
Wire Backward Hat Capital's $600 Million 'Audited' Forex Fund: What Manager Due Diligence Actually Missed NOV 19

Feature Data Privacy

GM Paid $220 Million for Consent Its Dealers Captured. Your Consent Management Has the Same Hole.

The 36-state GM settlement turns delegated consent capture and vendor data assessments into enforcement items. Consent management controls for dealer, agent, and partner channels.

Table of Contents

TL;DR

  • On October 29, 2026, New York Attorney General Letitia James and a 36-state bipartisan coalition announced a $220 million settlement with General Motors over selling driver data to brokers who resold it to insurers.
  • The mechanics matter more than the number: consent was captured by dealership staff in a combined signup flow, under enrollment bonuses and dealership rankings, and the insurance purpose was never disclosed.
  • The settlement’s remedies are program-level — an in-house data and privacy assessment program, a third-party privacy assessment program, and dealer oversight and training. That is the shape of every delegated-channel consent problem in financial services.

Most consent failures are not design failures. They are delegation failures. The privacy team writes a clean notice, product builds the modal, and then someone in a channel with a quota captures the click.

The General Motors settlement filed as Washington v. General Motors LLC et al. is the clearest enforcement artifact yet for that pattern, and it lands on financial services harder than it looks. Auto lenders, indirect auto programs, agent networks, ISOs, and partner branches all run consent through someone whose compensation does not depend on consent quality.

What the coalition alleged

GM’s Smart Driver feature, marketed since 2016 as a way to improve your own driving, collected trip counts, start and end times, hard braking and rapid acceleration frequency, and speed-limit exceedances. GM sold that data to data brokers, whose contracts specified insurance use, and the brokers resold it to insurers that priced premiums against it.

The company told drivers the data would not be shared with insurers without express consent, and never disclosed the insurance purpose. More than 700,000 New York drivers were enrolled. The New York Attorney General’s release describes a New York Times investigation in April 2024 that documented premium increases traceable to broker-supplied driving data, including one driver quoted a 22 percent increase and handed a 130-page report of six months of trips.

Three facts from the release do the real damage:

  • The signup was bundled. A single combined flow enrolled drivers in Smart Driver and OnStar together — directions, security, roadside assistance, and data sale, one acceptance.
  • Dealership staff accepted terms on customers’ behalf, routinely, at the point of purchase.
  • Enrollment was incentivized. From 2020 to 2024, GM paid dealer employees bonuses tied partly to enrollment volume and ranked dealerships on signups.

California ran a parallel case that produced a $12.75 million penalty against GM in May 2026, the largest CCPA penalty on record at the time — covered here when it landed. The multistate action packages the same facts into a program-remediation decree.

Then the app layer failed. GM’s OnStar app told more than 370,000 drivers they were not enrolled while their data was being collected and sold, and GM later quietly unenrolled them — without disclosing the error, and without getting the data back from brokers and insurers.

Why the incentive detail is the whole case

A salesperson at a dealership has a customer in front of them, a tablet, and a combined consent screen. The customer wants to drive off. Tapping through is the path of least resistance for both of them. If the salesperson’s bonus depends on enrollment, the tap gets faster.

Substitute terms and the same paragraph describes an indirect auto program, a mortgage broker, an insurance agent, or a partner branch onboarding deposit customers onto a co-branded app. The consent layer sits with someone who is measured on volume, trained quarterly at best, and employed by a third party.

Regulators have now priced that structure at $220 million. The remedy list maps straight onto it:

  1. Limit the data collected
  2. Build a comprehensive privacy program
  3. Establish an in-house data and privacy protection assessment program
  4. Establish a third-party data protection and privacy assessment program
  5. Add dealership oversight and training to prevent consent-free enrollment

Note what is not on the list: a better notice. Better notice was assumed. What the states demanded was an assessment regime and channel supervision.

Consent architecture in a delegated channel needs to record more than a boolean. An example capture record, as a starting schema rather than a form:

  • Actor: the individual who tapped, and whether they are the consumer or a staff member attesting on the consumer’s behalf, with the staff ID
  • Screen: the exact text rendered, versioned, plus a screenshot hash stored with the record
  • Scope: which purposes were accepted and which were not — separate toggles for service delivery, analytics, and third-party sharing or sale
  • Bundling disclosure: a field stating whether acceptance was conditionable on service use, and what the consequence of declining was
  • Timestamp and channel: device, location code, and channel type
  • Confirmation: what the consumer received afterward, and whether it listed the third-party categories receiving data

Two design rules follow. First, unbundle anything that leaves the building. If declining the data-sale leg kills the whole enrollment, you have a bundled consent and you should be able to say so on the record. Second, the consumer’s own receipt has to name the recipient categories, because that is the document a regulator will ask for when a customer says they never agreed to insurance underwriting.

Then test it. A quarterly mystery-shop of 10 to 15 enrollments per high-volume channel, scored against the script, with failure rates reported to the channel manager and to compliance. Ten to fifteen is a starter range: calibrate against your own first two cycles, and add channels where the rate runs hot. For the digital half of the same problem — opt-out signals firing on a marketing page while the enrollment flow ignores them — the GPC scoping rules cover which states require honoring the signal.

The two assessment programs, in practice

The settlement requires separate in-house and third-party assessment programs. Read that as two workstreams with different owners and different test procedures.

In-house: a documented assessment before new processing or a material change — purpose, data categories, recipients, retention, consumer notice, and the risk rating. California’s regulations already require roughly this structure for qualifying processing and require production to the agency on request. The program-level addition is a register, a review cadence, and a sign-off chain, so the assessments exist as a body of work rather than a folder of one-offs.

Third-party: an assessment of the recipient, not the relationship. For each broker, consumer reporting agency, analytics vendor, or data marketplace, the file should hold what data categories they receive, the contractual use restriction, whether the contract names insurance or underwriting use, the onward-transfer terms, and the deletion or return rights. The GM facts show why the contract clause alone is insufficient: GM’s contracts did specify insurance use, and the failure was that consumers were never told. The clause protected GM’s rights. It did not protect the disclosure.

Connecticut’s August 19, 2026 settlement with TaxAct shows where the third-party side is heading: a review committee approving new tracking technology, a data-point inventory for each tag, automated tag monitoring, and two independent audits of the tracking program. $275,000, and the audit obligations are the expensive part.

For financial institutions, tag-level governance is already unavoidable because of pixel litigation. The GM case adds the offline channel to the same framework.

Channel oversight that is not a training deck

Training gets a line in every settlement and almost never fixes the problem, because the failure is structural: staff are asked to explain a data sale during a handoff they are measured on. Three changes work better than a module.

  • Score the channel on consent quality, not just volume. Add a consent-accuracy metric to the dealer or agent scorecard, next to funded volume. What appears on the scorecard gets managed.
  • Put the decline path in the script. Staff need a sentence for “do I have to?” and it cannot be “it’s required.”
  • Reconcile enrollment to servicing. The 370,000-driver glitch was a state mismatch between app and back end. Reconcile enrollment state between the capture system and the servicing system monthly, and treat drift over threshold as an incident with a named owner.

A quiet unenrollment is worse than a visible one, because the customer’s belief and the data’s location diverge and only the broker knows the truth.

Where the internal argument happens

Expect pushback in two places.

Sales will argue that unbundling tanks conversion. Real, and usually overstated — the fix is to keep the flow to two screens and make declining a single tap, not to remove the ask.

Legal will argue that the GLBA exemption covers it. Partly. The carve-out is record-level, and plenty of what a modern program collects sits outside it: website and app telemetry, prospect data, and, for auto lenders, vehicle data that a dealer captured before the bank had a relationship. The GM theory was disclosure and purpose, and neither is cured by a safeguard-based exemption.

The position to take internally: every consent captured by a non-employee is an inference about what a customer understood. Your job is to make the record good enough that the inference holds.

So what

$220 million buys a privacy program. You can build the same three pieces — versioned consent capture, a third-party assessment file per recipient, and a scored channel review — for the cost of a quarter’s attention, and you will need them regardless of what happens in the auto sector.

Start with the recipient inventory: every vendor receiving customer data, the contractual use restriction, and whether your notice names that use. The gap list is your road map. Our Data Privacy Compliance Kit has the assessment template, the vendor terms review, and the applicability analysis by state.

FAQ

What did the states allege against General Motors?

The New York Attorney General, leading a 36-state bipartisan coalition, alleged that GM’s Smart Driver program collected detailed driving and location data and sold it to brokers under contracts specifying insurance use, while marketing the feature as a self-improvement tool and telling drivers data would not go to insurers absent express consent. The complaint is Washington v. General Motors LLC et al.

How much is the settlement and what did GM agree to?

GM agreed to pay $220 million, with New York’s share at $12,054,341, and to limit data collection, build a comprehensive privacy program, establish in-house and third-party data protection assessment programs, and add dealership oversight and training.

Why does a car company settlement matter to a bank?

The failure mechanism is channel delegation. Consent was captured by staff inside a bundled flow, under volume incentives, and data moved under contracts describing a use consumers never heard about. Banks run the same structure through dealers, agents, ISOs, and partner branches — and indirect auto lenders inherit dealer-captured consent directly.

When is a data protection assessment required?

Several state laws require a documented assessment before heightened-risk processing such as sensitive data, selling, or profiling. California’s regulations require identifying purpose, data categories, recipients, and consumer notification, and producing the assessment to the agency on request. The settlement adds a program obligation on top of case-by-case assessments.

Does GLBA exemption remove the risk?

Not entirely. The carve-out is record-level, and data captured outside GLBA scope — web and app telemetry, prospect data, dealer-captured vehicle data — can remain in scope. The enforcement theory was disclosure and purpose, which a safeguards-based exemption does not answer.

◆ Need the working template?

Start with the source guide.

These answer-first guides summarize the required fields, evidence, and implementation steps behind the templates practitioners search for.

◆ Immaterial Findings · Weekly

Sharp risk & compliance insights. No fluff.

◆ FAQ

Frequently asked questions.

What did the states allege against General Motors?
The New York Attorney General, leading a 36-state bipartisan coalition, alleged that GM's Smart Driver program collected detailed driving and location data from enrolled vehicles and sold it to data brokers under contracts specifying insurance use, while marketing the feature as a self-improvement tool and telling drivers their data would not go to insurers absent express consent. The complaint is Washington v. General Motors LLC et al.
How much is the settlement and what did GM agree to?
GM agreed to pay $220 million, of which New York's share is $12,054,341, and to limit the driving data it collects, build a comprehensive privacy program, stand up an in-house data and privacy protection assessment program and a separate third-party data protection and privacy assessment program, and add dealership oversight and training.
Why does a car company settlement matter to a bank or fintech?
The failure mechanism is channel delegation. Consent was captured by dealership staff inside a bundled signup flow, under enrollment-volume incentives, and data moved to vendors whose contracts described a use consumers were never told about. Banks run the same structure through dealers, agents, ISOs, and partner branches, and indirect auto lenders inherit dealer-captured consent directly.
What is a data protection assessment and when is one required?
Several state privacy laws require a documented assessment before processing that presents heightened risk — sensitive data, selling, profiling. California's regulations require the assessment to identify the purpose, the data categories, the recipients, and how consumers are notified, and to be produced to the regulator on request. The GM settlement adds a program-level obligation on top of case-by-case assessments.
Does GLBA exemption remove the risk for a bank?
No. The GLBA carve-out operates at the record level, and data captured outside GLBA scope — website behavior, app telemetry, prospect data, and in many setups vehicle or device data — can still be in scope for state law. The enforcement theory here was disclosure and purpose, not a missing safeguard.
Rebecca Leung

Author

Rebecca Leung

Rebecca Leung has 8+ years of risk and compliance experience across first and second line roles at commercial banks, asset managers, and fintechs. Former management consultant advising financial institutions on risk strategy. Founder of RiskTemplates.

◆ Related framework

Data Privacy Compliance Kit

Which of the 23 state privacy laws apply to your fintech after GLBA, plus the GLBA checklist, request tracker, assessments and vendor terms to comply.

Immaterial Findings · Newsletter

The brief, in your inbox.

Enforcement of the week, a framework breakdown, and the prompts that are actually worth running. Delivered to your inbox. Free.