Member
November 12, 2026

Assurance in Software Based Advertising Context Need and Foresight

Srijan Sanchar foresight note · 12 November 2026 · Shailendra Jaiswal

As advertising and content come to be decided by AI agents, a new function becomes necessary: independent assurance that receivers are not tracked beyond consent, children are not influenced, and algorithms behave as they claim. This is wider than cybersecurity. Cybersecurity protects systems from attackers; receiver assurance protects people from legitimate actors who stay inside the system and still cause harm. We expect it to become a major function by about 2030 (high confidence). Whether it stays independent and person-centred, or is captured by platforms, is still open (medium confidence).

1. Context: why an assurance gap is opening

Four shifts in software-based advertising remove the human from the point of decision, so trust can no longer be checked by looking.

• Agents decide out of sight. AI agents now buy, sell and place ads machine to machine. One seller alone reports over 25 trillion predictions a day.

• Content and advertising are merging. Fragments, serial ads, AI-inserted placements and brand-made content blur what counts as an ad.

• Orchestrators control access. Feeds, OS assistants and personal agents decide what reaches the receiver, and they capture most of the value.

• Value now depends on trust. Value of a moment ≈ attention × capability to purchase × trust; trust has to be produced and proven at machine speed.

The law is moving, but compliance on paper is not protection in practice.

Instrument	What it requires	Status
India: DPDP Rules, 2025	Verifiable parental consent for children’s data (Act s.9, which also bars tracking, behavioural monitoring and targeted ads at children); registered consent managers; due diligence on algorithmic risk by significant data fiduciaries	Notified 13 Nov 2025; consent-manager rule from 13 Nov 2026; fiduciary and children’s obligations from 13 May 2027
EU: Digital Services Act, Art. 28	Platforms may not show profiling-based ads to users they know to be minors	In force

Evidence of the gap. A 2026 algorithmic audit of TikTok found formal compliance with the EU rule, yet promotional content still reached minors with profiling 5–8 times stronger than formal ads to adults, strongest in undisclosed commercial content (KInIT, ACM FAccT 2026). The protection was escaped by moving persuasion from “ads” into “content”, exactly the convergence this foresight describes.

An assurance industry already exists, but it protects the advertiser. Verification firms check that ads are seen by humans and kept away from unsafe placements. DoubleVerify earned $748M in 2025, up 14% (results); Integral Ad Science earned $530M in 2024 (Digiday). No comparable industry yet assures the receiver.

2. Need: what must be assured

Receiver assurance must cover eight domains; cybersecurity covers only part of the first.

#	Assurance domain	The question it answers	Example of failure
1	Privacy and consent	Was data used only within the consent given, and did withdrawal take effect?	Intent shared for one purchase reused to profile across apps
2	Child safety	Were children kept out of profiling and persuasion, in ads and in content?	Undisclosed brand content profiled to minors (the TikTok audit)
3	Disclosure	Was every paid, sponsored or AI-inserted placement labelled?	AI-generated product placement inside a story, unlabelled
4	Agent conduct	Did the agent act for its principal (best execution) and leave a trace?	A buying agent routes spend to its owner’s exchange
5	Orchestrator neutrality	Did the gate favour its owner’s content or ads without saying so?	An assistant ranks its parent company’s offers first
6	Non-manipulation	Were vulnerability signals (grief, debt, illness, addiction) left unused?	Serial ads timed to late-night distress
7	Fairness	Did ad load, offer quality and prices stay fair across income and language?	Poorer or vernacular users see more ads and worse offers
8	Provenance	Is the content and its claims authentic and traceable to a source?	Deepfaked endorsement in a serial ad

Why this is wider than cybersecurity.

	Cybersecurity	Receiver assurance
Protects	Systems and data	People, especially children, and their attention
Against	Attackers outside the rules	Legitimate actors inside the rules
Asks	Was it breached?	Was it used rightly?
Evidence	Logs, intrusion detection	Decision traces, consent records, audits of what receivers actually saw
Buyer	The system owner	Receivers, parents, regulators, and brands protecting their reputation

Why it is needed now, through Srijan Sanchar frameworks.

• IEF constraint field: the foresight’s own constraints, consent kept revocable (C4) and every autonomous decision leaving a reviewable trace (C5), cannot be met without someone verifying them.

• CRRAAAS gates: the Trust gate turns from a feeling into an institution once the Agent gate sits between sender and receiver.

• Structural Threshold Foresight: legitimacy decay is at stage 3 of 4, and systems in that state respond by building assurance.

3. Foresight: how receiver assurance evolves to 2035

Assurance moves from protecting the advertiser to protecting the receiver, in three phases. Timings are Srijan Sanchar judgements.

Phase	Period	What happens	Trigger
1. Compliance assurance	2026–28	Consent-manager audits, children’s-data checks and DPIA-style reviews, driven by DPDP and DSA deadlines; done mostly as paperwork	DPDP obligations from May 2027; DSA enforcement
2. Behavioural assurance	2028–31	Audits test what receivers actually saw (as the TikTok audit did), not just policies; decision-trace checks on agents; labels for AI-inserted placements	Scandals in which compliant platforms still profile children; agent-buying disputes
3. Continuous, machine-speed assurance	2031–35	Assurance agents watch other agents in real time; devices attest to what on-device AI did; an assurance mark travels with each fragment and ad	Agent-to-agent trade becomes the default; OS assistants hold the gate

Three ways it could settle.

1. Captured assurance. Platforms pay their own auditors and certify themselves. Assurance becomes theatre that props up the walled model. Structural Threshold Foresight warns this is the most likely “intervention that suppresses volatility”: it delays the threshold without changing who controls attention.

2. Regulatory assurance. The state sets standards and licensed auditors check them, like financial audit. This is solid but slow, and weak on fast-changing AI behaviour.

3. Independent, person-centred assurance. Assurance answers to receivers and parents, uses open standards, and is paid for by a mix of regulation, brands and receivers’ agents. Only this outcome keeps Integrity, Impact and Vitality in balance (HEM).

Three conditions decide which path wins.

• Liability: penalties that bite under DPDP and the DSA.

• Verifiability: decision traces, consent records and device attestation, without which on-device AI cannot be audited.

• A payer independent of the audited: pooled industry funds, regulator-appointed auditors, or brands paying for receiver-side marks.

Where it sits in the value chain (Smile Curve). Receiver assurance becomes a new right-end activity. It sits beside orchestration and relationship, because whoever can prove trust earns access to qualified attention.

What history suggests (isomorphisms). Financial audit followed stock markets, food-safety certification followed packaged food, and cloud certification followed outsourced data. In each case assurance became a profession, with standards, licensed practitioners and a visible mark. Receiver assurance is likely to follow the same path.

4. Risks, Dharma guardrails and the opportunity

Risks to design against.

• Assurance theatre: marks awarded on policy documents, not on what receivers actually experienced.

• Captured auditors: the audited party chooses and pays the auditor.

• Assurance as a new toll: certification so costly that only large platforms afford it, excluding small Indian creators and MSMEs.

• Surveillance to prove safety: verifying age or consent by collecting more personal data than the harm it prevents. The EU rule deliberately does not require extra data to detect minors.

• Label blindness: so many marks that receivers stop noticing them.

Dharma guardrails, built before misuse.

• The auditor answers to the receiver, not the audited.

• Assure behaviour, not paperwork: test what people actually saw.

• Children are protected in content as well as ads.

• Proving safety must never require more data than the harm it prevents.

• Small creators get affordable, proportionate assurance.

• Every assurance mark is revocable, with the reason published.

The opportunity for Srijan Sanchar. Receiver assurance turns Srijan Sanchar’s Dharma gate from an internal principle into a category it can lead: Dharma Assurance.

Action	What it is	Intentionality (Thought–Action Network)	When
Publish the thought-leadership note	Name the category: receiver assurance, wider than cybersecurity	Committed	Now
Draft a Dharma Assurance standard	Eight domains as testable criteria, with behaviour-based tests	Provisional	Q1 2027
Child-safe assurance pilot with TAI	Certify toys and children’s content that collect no child data and use no persuasion aimed at children	Exploratory	2027, ahead of DPDP’s May 2027 obligations
Katha Setu self-assurance	Apply the standard to its own stories, labels and sponsorships	Exploratory	2027
Convene auditors, consent managers, regulators and child-safety experts	Shape an independent, receiver-paid assurance model	Provisional	2027

Claims Srijan Sanchar can stake: assurance is wider than cybersecurity; the next assurance industry protects the receiver, not the advertiser; children need protection in content, not only in ads.