Feature Data Privacy
Consent Receipts for Sensitive-Data Sales Must Survive Five Years. Most Records Are Two Fields.
Delaware HB 380 requires a five-year consent record produced with the DPIA. Here is the field set, the versioning rule, and the deletion conflict.
Table of Contents
TL;DR
- Effective January 1, 2027, § 12D-106(a)(12)(d) of the Delaware Personal Data Privacy Act requires a controller to maintain a record of consent to a sensitive-data sale for five years — and to produce that record with the related data protection assessment.
- ISO/IEC TS 27560:2023 consent receipts require only a receipt ID and a schema version. That proves a record exists; it does not prove the consent was informed. Delaware attaches to the record.
- The hard part is not the schema. It is versioned notice text, a revocation channel at least as easy as consent, and a written position on the five-year clock versus deletion.
The sentence is short and it sits at § 12D-106(a)(12)(d): “The controller maintains a record of the consumer’s consent for a period of 5 years.” The next subsection makes the record travel — it “must be provided with any data protection assessment produced under § 12-108(c) of this title.”
That is a pair of obligations most consent stacks cannot satisfy. Not because the fields are exotic, but because a five-year evidentiary chain needs three things teams rarely build: an immutable notice snapshot, a stored consent event with its method, and an owner who agreed in writing that the record outlives the customer.
What the Statute Actually Requires, In Order
Read the full text at 6 Del. C. ch. 12D and the HB 380 amendments as engrossed and the requirement is a sequence, each element load-bearing for the next:
- The disclosure of sensitive data is strictly necessary to provide or maintain a product or service the consumer affirmatively requested.
- Before the sale, a clear and conspicuous notice naming the specific sensitive-data categories, the purpose, and the third parties receiving the data.
- The consumer’s consent.
- A record of that consent, kept five years.
- The record produced alongside the § 12D-108(c) assessment.
Note what element four implies about elements two and three. If the notice changes — a new recipient, a broader purpose, a reworded category — the consent that followed the old notice is evidence of a different transaction. A record that points at “our current privacy policy” is unfalsifiable, because the policy will have been rewritten by the time anyone checks.
The sensitive-data definition is already wide, and it widened further this session: neural data from central nervous system activity now sits alongside health, genetic, biometric, precise geolocation and known-child data at § 12D-102(30). The neural-data signal test covers how to decide whether a given feed lands in that list.
Why a “Consent Receipt” Is Not the Artifact
The industry vocabulary pushes teams toward the wrong build. ISO/IEC TS 27560:2023 is the standard to know, and its own structure tells you the limits. As the W3C Data Privacy Vocabularies guide to implementing 27560 explains, a conformant receipt requires only a metadata section — a receipt identifier and a schema version. The field names in bold in the standard’s own figure are the mandatory ones, and the guide is explicit that the receipt “contains only two required fields.”
Two required fields will not survive a production letter. What does is the record, which 27560 organizes into four sections: header, processing, parties, events. Adopt that skeleton, then populate it past the standard’s minimum to meet Delaware.
The Record Field Set
Build once, render several outputs from it: the internal record, the consumer receipt, and the assessment appendix.
| Field | What it holds | Why it exists | Failure mode if skipped |
|---|---|---|---|
record_id | UUID-4, unique per consent event | 27560 § 6.3.3.3 requires a unique reference; it is how every other artifact points here | Records stitched together by email search instead of ID |
pii_principal_id | Account or pseudonymous identifier | 27560 § 6.3.3.4 requires association to a principal | Consent exists with no way to tie it to a person |
controller_entity | Legal entity, DE nexus, contact | Named party for any inquiry | Parent brand recorded, regulated entity unidentifiable |
notice_snapshot | Immutable copy of the exact notice text plus SHA-256 hash | Proves what the consumer saw | Record cites a policy URL that now resolves to different text |
notice_published_at | Timestamp the version went live | Establishes notice preceded consent | No ordering evidence; consent may postdate the notice |
sensitive_categories[] | Codes mapped to § 12D-102(30) | The notice must name categories | ”Sensitive data” recorded as one undifferentiated blob |
recipients[] | Legal names of third parties receiving the data | The notice must identify them | ”Partners and service providers” recorded instead |
necessity_rationale | The affirmatively requested product and why disclosure is strictly necessary | Satisfies element one | No answer when asked why the sale was permitted at all |
consent_event | Type, UTC timestamp, method (toggle, checkbox, signature), UI surface, locale | The consent itself | A boolean with no method or timestamp |
revocation_event | Type, timestamp, channel used, effective date | Connecticut § 42-520(a)(6) demands revocation at least as easy as consent, cessation within 15 days | Revocation by email only, when consent was an in-app toggle |
assessment_ref | assessment_id of the § 12D-108(c) DPIA | The statutory pairing | Five years of records with no assessment to pair them to |
retention_expiry | Consent date + five years, computed and stored | Forces the destruction decision to be deliberate | Records kept forever or deleted at contract renewal |
Two fields deserve their own paragraph.
notice_snapshot is the crux. Hashing alone gives you integrity without content — a hash proves the text has not changed, not what it said. Store both. The snapshot should be the rendered notice as served, including the category list and recipient names, with the hash computed over it. When counsel asks in 2029 whether the consumer was told about a specific recipient, the answer is a file, not a git log expedition.
revocation_event is where the model breaks. § 12D-106(a)(7) requires an effective revocation mechanism “at least as easy as the mechanism by which the consumer provided the consumer’s consent,” with processing to stop within 15 days of the request. Connecticut writes the same rule at § 42-520(a)(6), and it is the strictest phrase in either statute. If consent arrived through an in-product toggle, an email-only opt-out is not “at least as easy.” Record the channel used to grant consent, then make that same channel available for revocation, and log both with the same fidelity.
The Deletion Conflict Nobody Has Written Down
The five-year clock and a deletion request will collide in month three. The DPDPA’s § 12D-104(c)(5) carve-out allows a controller to satisfy a deletion request while retaining a record of the request and the minimum data needed to keep the data deleted. That is the hook to hang a position on, and it is a hook, not a rule — the statute does not resolve the interaction with the consent-record obligation.
Write the position now, before the first ticket. Something in the shape of: the consent record is retained for its statutory period in pseudonymized form; pii_principal_id is replaced with a salted hash; the record notes the deletion-request ID; no marketing use continues. Then get counsel to sign it, because the alternative is a support agent deciding in the moment whether to delete a record that the company is separately required to keep.
Same treatment for the expiry date. Five years from what — the consent event, the last sale under that consent, or the end of the relationship? The statute does not say. Pick the consent event as the conservative anchor, record the derivation, and revisit if the DOJ’s cure-factor analysis or the first enforcement action provides better guidance.
Prove It With Three Consumer Files
An evidence architecture you have never exercised is a hypothesis. Once a quarter, pick three consumers at random and try to produce a complete packet: record, notice snapshot, assessment, recipient list, and revocation history. Time it. If a competent analyst cannot assemble all three inside a working day, the schema has a hole and you have found it for free.
Then run the reverse test. Take one record at random and re-derive the consumer’s current consent state from it. If the record cannot tell you whether consent is live, revoked, or superseded by a newer notice, it is a filing cabinet, not a record.
Keep the results. Under § 12D-111(c) the Department of Justice weighs several factors when deciding whether to allow a cure, and one of them is “whether such alleged violation was likely caused by human or technical error.” Quarterly test logs are how you distinguish a design flaw being worked from a systemic gap being ignored.
Retention and Destruction Rules
- Anchor expiry to the consent event, not the contract or the account closure.
- Destroy in a scheduled job at expiry, but gate on two conditions: no open assessment referencing the record, no litigation hold.
- Never destroy the record while the paired DPIA is still live — the statutory pairing means the record’s absence breaks the assessment.
- Log every destruction with
record_id, date, and authority, so a later request gets a dated answer instead of a shrug. - Keep the notice snapshot until the record dies. A record surviving its notice is only half evidence.
So What?
Most teams will ship a consent receipt this quarter, check “done,” and discover in 2029 that they retained a receipt ID and a schema version with no notice text behind it. The build that survives is unglamorous: one immutable record per consent event, a versioned notice hash, a revocation channel with equal friction, and a signed position on deletion. Delaware’s January 1, 2027 date gives you six weeks to install it and four years to be glad you did.
The Data Privacy Compliance Kit includes the consent record schema, notice versioning procedure, deletion-conflict memo template, and DPIA cross-reference register. For scope questions, the HB 380 applicability walkthrough covers the threshold math and the GLBA exemption boundary.
FAQ
Do we need a receipt for consent we already hold? The obligation attaches to sales on or after January 1, 2027. Pre-existing consent can often be reused if it was informed and the notice still describes the current recipients and purpose. Where it does not, re-paper before the first sale rather than retrofitting a record you cannot support.
Can a consent management platform solve this? Only if it stores notice versions immutably and exposes the full event trail. Many CMPs store a consent string and a policy ID. Ask to see a rendered historical notice for a specific consent before you sign.
What if the third party changes its legal name? Amend the notice version and record the successor against the existing recipient entry. A name change on a live recipient is a notice event, because the consumer was told a specific entity was receiving the data.
Does this apply to processor disclosures, or only sales? The five-year record is a sensitive-data sale obligation under § 12D-106(a)(12). Consent to sensitive-data processing generally sits at § 12D-106(a)(4), and in practice one record serves both. Build one schema and tag the event type.
◆ 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.
◆ Related template
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 · Weekly
Sharp risk & compliance insights. No fluff.
◆ FAQ
Frequently asked questions.
What does Delaware HB 380 require for a sale of sensitive data?
Is a Kantara or ISO 27560 consent receipt enough on its own?
How does the five-year clock interact with a consumer deletion request?
Who is actually in scope for this?
Does the record have to be given to the consumer?
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.
◆ Keep reading
Related posts.
Data Privacy
Neural Data Becomes Sensitive Data in Delaware on January 1. Run the Signal Test First.
Delaware HB 380 makes neural data sensitive data on Jan 1, 2027, joining California, Colorado, Connecticut and Montana. Here is the measured-vs-inferred test and the vendor evidence.
Nov 16, 2026
Data Privacy
Delaware Just Narrowed the GLBA Privacy Exemption. Run the Applicability Test Before January 1.
Delaware HB 380 takes effect January 1, 2027, cuts the threshold to 10,000 consumers, and strips the entity-level GLBA exemption from fintechs. Scope it now.
Nov 14, 2026
Data Privacy
Texas's AI Law Changed What Your AI Vendors Owe You. TRAIGA's TDPSA Amendments Have Been in Effect Since January.
The Texas Responsible AI Governance Act (TRAIGA, HB 149) took effect January 1, 2026, amending the TDPSA to add AI system processor obligations. Federally insured financial institutions get a carve-out from TRAIGA's non-discrimination rules — but not from the processor amendment. Here's what fintechs and banks using AI for Texas customers need to assess now.
Oct 11, 2026