Liability for Scam Adverts

Require platforms to take reasonable steps against fraudulent adverts, and remove their immunity for the ones they are paid to run.

AI evaluation · not yet reviewed by a human

This evaluation was produced and sourced by an AI model; a human review is still pending. Figures and conclusions may still change. The review log is at the foot of the page.How review works →

A platform that hosts what a user posts is shielded from liability by section 230 of the Communications Act. The same shield currently covers advertisements the platform was paid to place, targeted and delivered by its own systems. The bipartisan bill before Congress would impose a duty to take reasonable steps against fraudulent and deceptive adverts — verifying advertisers, acting on reports, keeping a record — and would remove the section 230 defence for paid content that breaches it. Organic posts, messages and everything a user writes are untouched. This evaluation looks five years ahead.

Balance

Better for the future · 0.67 previous scale

Balance on the previous scale. The Bilanz 2.0 simulation is not yet available for this evaluation. The category comes from the share of the debate on the pro side (r).

For 27 · 67 % Against 14 · 33 %

Size class: medium Scale of this evaluation: Normalised Impact — unitless, calibrated to this topic. For comparison: one point here is worth roughly 200 million euro per year. Reported fraud losses are multiplied by four here to reach actual losses, because the trade commission's own work suggests only a small minority of victims report. At the reported figure alone this measure barely registers; at the multiplier its own research implies it would be twice the size shown. How we score →

Arguments for

Arguments against

6 arguments evaluated · Scoring v1.3 Δ absolute +13

Arguments — For

3 arguments

Money that stops leaving the country

15of 100

Americans reported losing 2.1 billion dollars in 2025 to scams that began on social media, eight times the 2020 figure. A large share of it starts with an advertisement the platform was paid to deliver. The money goes to organised fraud operations abroad and does not come back.

Value 5 · Household budgetsImpact 6.0Plausibility 5

▸ Show reasoning & sources ▾ Hide reasoning & sources

The stream is money, priced at the middle of the scale, but it is not a transfer in the ordinary sense. A payment obtained by deception is not an exchange either side chose on informed terms, and the recipients are almost entirely criminal operations outside the country, so nothing offsets the loss anywhere in this accounting. That is why the whole amount is counted rather than a difference in what a euro is worth at two ends. The distress of being defrauded is a separate stream and is counted separately, because losing four thousand euro to a fake investment is not the same event as spending four thousand euro. The value is the middle of the scale, and the whole sum is counted because there is no second end to the transfer inside this accounting.

Americans reported 2.1 billion dollars of losses in 2025 to scams that began on social media, eight times the 2020 figure, with about 1.1 billion of it in investment fraud ^[1]. How much of that starts with a paid advertisement rather than a message, a marketplace listing or an organic post is the first estimate: 35 percent is used, in a range from 15 to 60, since investment and shopping fraud — the two largest categories — are the ones most often advertised. The second and larger correction is under-reporting. The trade commission's own work suggests only a small minority of defrauded people ever file a report; a multiplier of four is used here, in a range from two to ten, which is well below what its own estimate implies. That gives about 2.53 billion euro of actual losses from paid scam adverts. A platform facing liability screens its advertisers, and 40 percent of the losses are assumed prevented, in a range from 15 to 70 percent. The victims carry a weight of 1.2. The Impact is the largest in this debate and it rests on two multipliers rather than on any direct count of scam adverts.

▸ Show calculation ▾ Hide calculation

	Reported losses to scams that began on social media [1]	2.1 billion dollars, eight times the 2020 figure	1,810 million euro
×	Share starting with a paid advertisement Setting, range 15 to 60 percent: investment and shopping fraud, the two largest categories, are the ones most often advertised [1]	35 %	634 million euro
×	Correction for losses never reported Setting, range 2 to 10: the trade commission's own work suggests only a small minority of defrauded people file a report, which implies a higher multiplier than the one used	× 4	2,530 million euro
×	Share prevented once platforms face liability Setting, range 15 to 70 percent: a platform that can be sued for a paid advert screens the advertiser	40 %	1,010 million euro
×	Weight of a euro at these incomes fraud losses fall across the income distribution and hit hardest relative to means below the middle of it	1.2	1,212 million euro
÷	Normalised Impact scale of this evaluation	200 million euro a point	6.06

Score 6.06 Impact × 5 Value × 5 Plausibility ÷ 10 = 15 of 100

The reported loss figure is solid and everything built on it is estimated. The counterfactual is the current position, in which platforms face no liability for adverts they are paid to run and enforce their own policies at their own pace. The trade commission's series is a census of complaints rather than a survey, which makes it exact about what was reported and silent about what was not, and the under-reporting multiplier is where most of the uncertainty lives. What the measure would achieve has no measurement at all: the United Kingdom imposed a fraudulent advertising duty in its online safety regime and Australia built a scams prevention framework, both too recent to have produced an evaluation. The confounder that matters is displacement — fraud that cannot be advertised moves to messaging, marketplace listings and organic posts, none of which this bill touches — and it is unresolved. Reverse causation does not arise. The Plausibility is at the middle: the losses are counted precisely, the share attributable to adverts and the share preventable are both assumed, and no comparable duty has yet been evaluated anywhere.

evidence basis: Projection · P ceiling 6 identification: Mechanistic · rung ceiling 6

Counterfactual: the current position, with no platform liability for paid adverts. Design: mechanistic — chain named (liability → advertiser screening → fewer scam adverts → fewer losses), with no evaluated precedent; the United Kingdom and Australian duties are too recent. Confounder: displacement of fraud into messaging, marketplace listings and organic posts, which the bill does not touch; unresolved. Direction: no reverse causation. Ceiling: projektion 6.0 binds and mechanistic gives the same. The under-reporting multiplier and the advert share are both carried in bands rather than in P.

What being defrauded does to people

10of 100

Losing a retirement account to a fake investment is not the same experience as spending it. Victims report shame that stops them telling anyone, damage to marriages, and in the worst cases they do not recover. That is a separate harm from the money and it is not small.

Value 9 · HealthImpact 2.5Plausibility 4.5

▸ Show reasoning & sources ▾ Hide reasoning & sources

The stream is what happens to a person after they discover they have been deceived: the depression and anxiety, the shame that keeps most victims from telling their family, the marriages that do not survive it, and at the far edge the suicides that follow the largest losses. This site places it in the class it uses for life and health, one step below the top because most people do recover. It is genuinely separate from the money, which is counted above: a household that loses the same sum in a bad investment it chose does not experience this. What is priced is the victim's own condition, not the effect on the people around them, which nothing here measures. The value sits one step below the maximum: the stream is health, and health that can be regained.

Behind about 2.53 billion euro of actual losses from advertised fraud sit roughly 630,000 victims a year, at a median loss in the low thousands. If the measure prevents 40 percent of the losses it prevents a similar share of the victimisations: about 252,000 people a year who are not defrauded. What that spares each of them is set at 0.05 quality-adjusted years, in a range from 0.02 to 0.15 — a few weeks of serious distress on average, which is a low reading given that the tail of this distribution contains people who lose everything they have. That gives 12,600 quality-adjusted years a year. Nothing is counted for the family members who carry it with them, and nothing for the smaller but far more numerous group who are targeted and not deceived. The Impact is under half the money it accompanies, which is the ordinary proportion when a harm is severe for a minority and moderate for most.

▸ Show calculation ▾ Hide calculation

	Actual losses from advertised fraud [1]	from the argument above	2,530 million euro
÷	Victims a year Setting, range 2,000 to 8,000 euro: the median reported loss sits in the low thousands	about 4,000 euro each	630,000 people
×	Share spared by the measure the same share as the losses prevented above	40 %	252,000 people
×	Quality-adjusted years spared each Setting, range 0.02 to 0.15: a few weeks of serious distress on average, which is low given that the tail contains people who lose everything	0.05	12,600 quality-adjusted years
×	Value of the years the value of a healthy life year used across this site	40,000 euro each	504 million euro
÷	Normalised Impact scale of this evaluation	200 million euro a point	2.52

Score 2.52 Impact × 9 Value × 4.5 Plausibility ÷ 10 = 10 of 100

That fraud victims suffer measurable psychological harm is documented in a substantial clinical and victimology literature, and the direction is not disputed by anyone. What that literature lacks is a counterfactual: it studies people who were defrauded, generally after the fact, against population norms rather than against a matched group who were targeted and escaped. Selection is therefore unresolved — people in difficult circumstances may be both easier to defraud and worse off afterwards for reasons that have nothing to do with the fraud. The victim count used here is derived from the loss figure and a median rather than counted, and the quality-adjusted figure is a construction of this evaluation with no source behind it. What holds the argument up is that its direction is beyond dispute and its size is deliberately set at the low end. Reverse causation is a live concern and is the reason for that choice. The Plausibility is below the middle: the harm is certain and every number attached to it here is assumed.

evidence basis: Mechanism · P ceiling 5.5 identification: Associational · rung ceiling 5.5 band: Chain closed, unevidenced · P 4–5

Counterfactual: population norms rather than a matched group of people targeted and not deceived — the victimology literature does not supply one. Design: associational — after-the-fact comparison without exogenous variation. Confounder: circumstances that make a person both easier to defraud and worse off independently; unresolved. Direction: reverse causation is live and is why the quality-adjusted figure is set at the low end of its range. Ceiling: associational 5.5 binds. Band: chain closed but unevidenced — the chain is named, the selection concern is answered by choosing a low figure, and only the measurement is missing.

Nothing measured argues against the claim; what is missing is any study comparing defrauded people against a matched group who were targeted and escaped. The selection concern is answered by setting the harm at the low end of its range. Read back: about half the time, being spared a fraud is worth roughly the amount of healthy time assumed here.

Open: Fraud reports carry contact details and platforms hold records of who saw which advert. A follow-up of reported victims against targeted non-victims would give the first clean estimate and could carry P to 6.

The advertising channel is worth more when it is clean

2.4of 100

Scam advertisers bid against honest ones for the same attention, which raises what everybody pays. They also teach users to distrust adverts generally, which lowers what the attention is worth. Both effects run against the businesses that pay for the platform.

Value 6 · OutputImpact 1Plausibility 4

▸ Show reasoning & sources ▾ Hide reasoning & sources

The stream is the efficiency of a market that moves about 117.7 billion dollars a year in the United States: advertisements reaching people who want the product rather than people who will be defrauded. It belongs to the class this site uses for economic systems and prosperity. What is counted is not the advertising spending, which moves from a company to a platform and stays inside the economy, but the difference between what the channel delivers when it is trusted and what it delivers when it is not. The platforms' own revenue is not counted as a gain or a loss, because a cleaner channel that carries fewer adverts at higher prices is roughly neutral for them. The value sits in the middle-upper part of the scale, at the level this site uses for economic output.

American social media advertising was worth 117.7 billion dollars in 2025, or 101 billion euro ^[3]. Fraudulent advertisers compete for the same inventory and can outbid honest ones because their expected return per impression is higher, which raises prices for everyone else, and their presence makes users warier of every advert they see. Both effects are small per unit and large in aggregate. A gain of 0.2 percent in what the channel delivers is used, in a range from 0.05 to 0.8 percent — 202 million euro a year. That figure is a construction; nobody has measured what a scam-free feed would be worth to the businesses buying it. What is not counted is the reputational value to the platforms themselves, which is theirs rather than anyone else's. The Impact is the smallest on this side, which is the honest scale of an efficiency gain in a market where fraud is a small share of the volume.

▸ Show calculation ▾ Hide calculation

	American social media advertising [3]	117.7 billion dollars at 1.16 to the euro	101 billion euro
×	Gain in what the channel delivers once fraudulent bidders are gone Setting, range 0.05 to 0.8 percent: fraudulent bidders raise clearing prices and make users warier of every advert; nobody has measured the size	0.2 %	202 million euro
÷	Normalised Impact scale of this evaluation	200 million euro a point	1.01

Score 1.01 Impact × 6 Value × 4 Plausibility ÷ 10 = 2.4 of 100

The mechanism is standard auction economics and the size has never been estimated for this market. The counterfactual is the same advertising market with scam advertisers still bidding. Two links are visible and one is not: that fraudulent bidders raise clearing prices follows from how the auctions work, and that users who distrust adverts respond to fewer of them is well established in advertising research, but nothing connects those to a number for this channel. The confounder that matters runs against the argument: platforms already remove large volumes of fraudulent adverts under their own policies, so the marginal effect of a legal duty on top may be small. That is unresolved. Reverse causation does not arise. The Plausibility is below the middle: the mechanism is uncontroversial, the size is invented, and the platforms' existing enforcement may already capture most of it.

evidence basis: Mechanism · P ceiling 6 identification: Mechanistic · rung ceiling 6 band: Chain closed, unevidenced · P 4–5

Counterfactual: the same advertising market with fraudulent bidders present. Design: mechanistic — auction mechanics and advertising response research support the direction, nothing supports the size. Confounder: platforms' existing voluntary enforcement already capturing most of the effect; unresolved. Direction: no reverse causation. Ceiling: mechanistic 6.0 binds. Band: chain closed but unevidenced — the links are named and the existing-enforcement counter-mechanism is stated; only the measurement is missing.

Nothing measured argues against the claim; what is missing is any estimate of what a cleaner channel is worth to advertisers. The counter-mechanism — that platforms already remove fraudulent adverts voluntarily — is stated and unresolved. Read back: about half the time, the channel gains roughly the efficiency assumed here.

Open: Advertising clearing prices by category are visible to large buyers. Comparing them before and after a platform tightens advertiser verification would put a number on the auction effect.

Arguments — Against

3 arguments

Honest advertisers get refused

8.1of 100

A platform that can be sued for a bad advert will reject anything that looks like one. That means new advertisers with no history, small businesses without documentation, anything about health or money, and adverts in languages the reviewers do not read. None of them are scams and all of them look like risk.

Value 6 · OutputImpact 3Plausibility 4.5

▸ Show reasoning & sources ▾ Hide reasoning & sources

The stream is business that does not happen: a shop that cannot reach customers, a service that never finds its market, a founder who gives up because the only channel available to them is closed. It belongs to the class this site uses for economic systems and prosperity. What is counted is the value of the transactions that do not occur, not the advertising spending itself, which simply goes somewhere else. Nothing is priced for the unfairness of being refused, which is a real grievance and not a separate good. The value sits in the middle-upper part of the scale, at the level this site uses for economic output.

The screening this bill forces is not costless in accuracy. Every system that reviews adverts at scale refuses legitimate ones, and the rate rises sharply when the reviewer bears the cost of a mistake in one direction and not the other. Two percent of legitimate advertising being wrongly refused is used here, in a range from half a percent to six percent: about 2.02 billion euro of the 101 billion euro market. The advertisers refused do not lose that spending — they spend it elsewhere or keep it — so what is lost is the surplus the advert would have generated above its cost, put at 30 percent, in a range from 15 to 60. That gives about 606 million euro a year. The loss concentrates among the advertisers least able to argue: new businesses, small ones, and those advertising in languages the review systems handle worst. The Impact is half the money the measure saves, which is the honest cost of asking a platform to judge which of its customers are criminals.

▸ Show calculation ▾ Hide calculation

	American social media advertising [3]		101 billion euro
×	Legitimate advertising wrongly refused Setting, range 0.5 to 6 percent: refusal rates rise when the reviewer bears the cost of a mistake in one direction only	2 %	2.02 billion euro
×	Surplus the advert would have generated above its cost Setting, range 15 to 60 percent: the advertiser keeps or redirects the spending, so what is lost is the business it would have brought	30 %	606 million euro
÷	Normalised Impact scale of this evaluation	200 million euro a point	3.03

Score 3.03 Impact × 6 Value × 4.5 Plausibility ÷ 10 = 8.1 of 100

That liability produces over-removal is one of the better-established regularities in platform regulation, observed wherever a takedown duty has been imposed with penalties on one side only. The counterfactual is the current position, in which a platform's incentive is to sell the advert. What has no measurement is the rate. Platforms do not publish wrongful refusal figures, no regulator collects them, and the two percent used here is a construction — the same figure could be justified anywhere between a fifth of that and three times it. The counter-mechanism is real and partly answered: the bill sets a reasonable-steps standard rather than strict liability, which is designed precisely to stop platforms refusing everything ambiguous, and how far a court would let that defence run is unknown. Reverse causation does not arise. The Plausibility is below the middle: over-removal is a well-observed response to one-sided liability and its rate here has never been measured.

evidence basis: Mechanism · P ceiling 6 identification: Mechanistic · rung ceiling 6 band: Chain closed, unevidenced · P 4–5

Counterfactual: the current position, in which a platform's incentive is to sell the advert. Design: mechanistic — over-removal under one-sided liability is a widely observed regularity, with no measured rate for advertising specifically. Confounder: the bill's reasonable-steps standard, which is designed to prevent blanket refusal; partly answered, since how a court would apply it is unknown. Direction: no reverse causation. Ceiling: mechanistic 6.0 binds. Band: chain closed but unevidenced — links named, the reasonable-steps counter-mechanism stated, only the rate unmeasured.

Nothing measured argues against the claim; what is missing is any published rate of wrongful advert refusal. The counter-mechanism — the reasonable-steps standard rather than strict liability — is named and its effect on courts is unknown. Read back: about half the time, a duty of this kind wrongly refuses roughly the share of legitimate advertising assumed here.

Open: Platforms hold appeal and reinstatement rates for refused adverts. Publishing them before and after the duty takes effect would measure over-removal directly and could carry P to 6.

Verifying every advertiser costs money

4.1of 100

Reasonable steps means knowing who is buying the advert: a business registration, a payment trail, a person to hold responsible. Doing that for the millions of advertisers on a large platform is a real operation, and it has to be redone as advertisers churn.

Value 5 · Enforcement costImpact 1.5Plausibility 5.5

▸ Show reasoning & sources ▾ Hide reasoning & sources

The stream is money spent on checking rather than on anything the checking produces: verification staff, identity services, appeal handling, record keeping. It is priced at the middle of the scale like any other money and is a genuine cost rather than a transfer, since the hours and the systems are consumed. Whether the platform absorbs it or passes it to advertisers makes no difference to the weight, and it will mostly be passed on. Nothing is counted here for the enforcement burden on regulators or courts, which is small beside the private cost. The value is the middle of the scale, the level this site uses for money spent on running a rule.

Large platforms carry something in the order of ten million active advertisers between them, most of them small and many short-lived. Verifying an advertiser to a standard that would survive a court — a registration, a beneficial owner, a payment trail — costs something like 30 euro a year each once appeals and re-verification are included, in a range from 10 to 90. That gives about 300 million euro a year. The figure excludes the review of individual adverts, which platforms already do at scale and would extend rather than build. It also excludes the cost of litigation itself, which is genuinely unpredictable: the value of removing a section 230 defence is that cases can proceed, and the first years of any such change are expensive in ways nobody forecasts well. The Impact is a quarter of the money the measure saves, which is the ordinary proportion for a duty that requires knowing your customer.

▸ Show calculation ▾ Hide calculation

	Active advertisers across large platforms [3]		10 million
×	Cost of verifying each to a standard that survives a court Setting, range 10 to 90 euro: a payment card check and a documentary beneficial-owner check differ by a factor of five, and the duty's standard is not yet settled	30 euro a year	300 million euro
×	Weight of a euro in company budgets the standard weight this site uses for business money	1.0	300 million euro
÷	Normalised Impact scale of this evaluation	200 million euro a point	1.5

Score 1.5 Impact × 5 Value × 5.5 Plausibility ÷ 10 = 4.1 of 100

Verification costs are known, because identity verification is a mature commercial service with published pricing, and the number of advertisers is reported by the platforms themselves. The counterfactual is current practice, in which advertiser verification is partial and voluntary. What is estimated is how far the standard would go — a reasonable-steps duty could be satisfied by a payment card check or could require documentary proof of a beneficial owner, and the difference between those two is a factor of five. That is why the range is wide. The confounder that would lower the figure is that the largest platforms already verify some advertiser categories under their own policies and under existing political advertising rules, so part of this cost is already being paid. Reverse causation does not arise. The Plausibility is at the upper end of what a projection can carry: the unit costs are commercial prices and only the standard they must meet is uncertain.

evidence basis: Projection · P ceiling 6 identification: Definitional · no rung ceiling

Counterfactual: current practice, with partial and voluntary advertiser verification. Design: definitional — a duty to verify requires verification, which has a commercial price; no behavioural link carries the quantity. Confounder: existing voluntary and political-advertising verification already covering part of the cost; named and unresolved. Direction: not applicable. Ceiling: projektion 6.0 binds because the standard the duty would require is uncertain by a factor of five, which sits in the 10 to 90 euro band.

Smaller platforms stop selling adverts

1.4of 100

The verification operation this requires has a fixed cost that a platform with a billion users barely notices and one with two million cannot carry. The predictable result is that advertising concentrates further with the four companies that already have most of it.

Value 6 · CompetitionImpact 0.6Plausibility 4

▸ Show reasoning & sources ▾ Hide reasoning & sources

The stream is competition in a market that is already concentrated: fewer places for an advertiser to go, fewer platforms able to fund themselves without being acquired. This site places that in the class it uses for economic systems and the working order of markets. What is priced is the lost competitive pressure rather than any particular company's survival, and no weight is given to smallness for its own sake. The compliance cost to the smaller platforms themselves is inside the previous argument and is not repeated here. The value sits in the middle-upper part of the scale, at the level this site uses for the working order of markets.

American social media advertising is already dominated by a handful of companies, and the fixed cost of a verification and appeals operation is exactly the kind that entrenches such a position. Smaller platforms have three options: build it, buy it from a vendor at a worse unit price, or stop selling advertising to anyone they cannot easily check. The value of the competitive pressure lost is put at 120 million euro a year, in a range from 25 to 400 million, which is roughly a tenth of a percent of the market. It is a price set rather than derived, and the reason it is small is that the concentration it worsens is already close to complete. The bill contains thresholds intended to spare the smallest operators, and how those are drawn decides whether this argument is worth anything at all. The Impact is the smallest in this debate and it is the one that a threshold in the bill's text would remove.

▸ Show calculation ▾ Hide calculation

	Value of the competitive pressure lost as advertising concentrates further Setting, range 25 to 400 million euro: roughly a tenth of a percent of a market already dominated by a handful of companies [3]		120 million euro
÷	Normalised Impact scale of this evaluation	200 million euro a point	0.6

Score 0.6 Impact × 6 Value × 4 Plausibility ÷ 10 = 1.4 of 100

The mechanism is standard and its application here is speculative. That fixed compliance costs favour incumbents is among the better-supported findings in regulatory economics, observed across banking, pharmaceuticals and financial services. The counterfactual is the current advertising market. What is absent is any estimate for this case, and the chain has an unresolved link: whether the duty would apply to a platform below a size threshold, which is a drafting question the bill answers with a definition that has not been finalised. The counter-mechanism is genuine — vendors sell advertiser verification as a service, and a small platform buying it at a per-check price faces no fixed cost at all, which would remove the argument entirely. That is not resolved. Reverse causation does not arise. The Plausibility is below the middle: the mechanism is well supported in general, the size is a stated price, and a vendor market may remove the fixed cost that the whole argument rests on.

evidence basis: Mechanism · P ceiling 6 identification: Mechanistic · rung ceiling 6 band: Chain closed, unevidenced · P 4–5

Counterfactual: the current advertising market. Design: mechanistic — fixed compliance costs favouring incumbents is well supported across other regulated sectors, with no estimate for this case. Confounder: a vendor market for advertiser verification removing the fixed cost entirely; unresolved and capable of eliminating the argument. Direction: no reverse causation. Ceiling: mechanistic 6.0 binds. Band: chain closed but unevidenced — links named, the vendor counter-mechanism stated, only the price unmeasured.

Nothing measured argues against the claim; what is missing is any estimate of the competitive value at stake. The counter-mechanism — that verification is sold as a per-check service with no fixed cost — is named and unresolved. Read back: about half the time, this duty entrenches the incumbents by roughly the amount assumed here.

Open: The size threshold in the final bill text settles most of this. Advertising revenue shares by platform, before and after the duty, would measure the rest.

Summary

This is the most conventional measure in the area and it comes out well because the harm it addresses is measured and the costs it imposes are ordinary. Americans reported 2.1 billion dollars of losses to scams that began on social media in 2025, eight times the 2020 figure, and the money goes to criminal operations abroad and never returns. The platforms are paid to deliver a large share of it and are currently shielded from any consequence, which is an unusual position for a business that takes money to place a message. What holds the balance back from being one-sided is that liability of this kind reliably produces over-caution: a platform that can be sued for a bad advert refuses new advertisers, small businesses and anything about health or money, and that cost lands on people who did nothing wrong. Two numbers decide the result and neither is measured — how much of the fraud actually starts with a paid advert, and how much of it a screening duty would stop.

Outlook — effect over time

Better for the future · 0.67 previous scale

Sources

Federal Trade Commission: Reported losses to scams on social media eight times higher than in 2020. ftc.gov
Congress.gov: S. 3774, Safeguarding Consumers from Advertising Misconduct Act. congress.gov
Interactive Advertising Bureau: Internet Advertising Revenue Report. iab.com

Last reviewed by Claude Opus 5 · September 6, 2026 · 2× AI, not yet reviewed by a human

September 6, 2026AI reviewClaude Opus 5record updated
i_spanne an allen 6 Argumenten aus den englischen Ketten, kein Transfer im Record. Kategorie kippt von Besser (r 0,67) auf Ausgeglichen (P(D>0) 0,69).
September 6, 2026AI reviewClaude Opus 5First evaluation
Created for the English side: reported fraud losses corrected upward for under-reporting, with the over-removal cost of one-sided liability booked against it.

Evaluations are produced with AI support and reviewed on a schedule for new developments; human passes are marked separately.How we review →