What It Should Cost a Stranger to Reach You

A stranger can send email for almost nothing per message, and that fact has shaped the economics of email for thirty years. A sender who pays nothing per message can mail a hundred million people and profit from a handful of replies, while the lost time, the filtering bills and the fraud losses land on everyone else. Large language models have now removed the last cost the old spammer still paid, which was writing something a person might believe. A message that once arrived with broken grammar now arrives with a colleague's name, a reference to last week's meeting and a plausible invoice attached.

The Accord's answer is to let every person put a price on their attention. A stranger who wants to reach you pays that price, to you, when the message is delivered. The sender sees the price before pressing send and decides whether the message is worth it. People you already correspond with reach you free. The Accord calls this attention postage: postage paid to the person whose attention is being taken, at a rate that person sets.

Attention belongs to the person who spends it

When the Accord prices a harm, the payment goes toward whoever bears it. Part of the carbon fee returns to households as the Energy Stipend. The Employer Parity Surcharge, which employers pay when they hire workers admitted from abroad, funds the local services that absorb those workers' families. An unsolicited message spends the recipient's time and concentration, so the recipient is the one who should be paid.

The right to refuse is settled law. In 1970 the Supreme Court upheld a statute under which any householder could have the Post Office order a mailer to stop sending advertisements that the householder, in his sole discretion, found sexually provocative. The Court wrote that "no one has a right to press even 'good' ideas on an unwilling recipient" and that "a mailer's right to communicate must stop at the mailbox of an unreceptive addressee" (Rowan v. Post Office Dept., 397 U.S. 728). No one is obliged to watch the evening news, however much it might do the country good. By the same reasoning, no one should have to lower their guard for a mass mailing because its cause is political or charitable. Attention postage treats every sender alike: a merchant, a campaign, a charity and a fraudster all see the same price. The law judges no content and sets no one's price.

Unpriced email has costs beyond fraud. In 2003 a quarter of email users told the Pew Internet & American Life Project that spam had reduced their use of email, and 52 percent said it had made them less trusting of email in general (Fallows, Pew Internet, 2003). I know people who have abandoned an address to escape the flood and lost touch with old friends in the process. A price lets a person keep one address for life.

How it works

For the person receiving mail, the change is a setting. Each account holder chooses a price for mail from strangers, anywhere from zero up to a ceiling set by their email provider. One provider might cap prices at a quarter; another might let a customer who guards their concentration charge $5. Providers compete on the ceilings they offer. The law requires only that every covered provider offer the mechanism, on both the sending and the receiving side.

Mail from people you already know arrives free. That includes anyone in your contacts, anyone you have written to and any list you subscribed to. When you give your address to a business, your provider can issue a permission address for that business, so receipts and order updates arrive free. Free status is yours to withdraw. A merchant that mails you several times a day after one purchase can be switched back to paying your price with one setting, or you can set any known sender to start paying after a set number of messages a week. The merchant then decides whether each message is worth what it costs.

Nothing is asked of the recipient when paid mail arrives, because the price did its screening before the message was sent. A recipient who wants to can turn on automatic refunds, so that replying to a stranger or saving the stranger as a contact returns the postage. A reply also makes the sender a correspondent, and later messages in either direction travel free.

For the sender, the price is visible before the message goes. The mail program looks up each recipient's price as the address is entered and shows the total before sending. The sender sets a maximum per recipient, and any address priced above it is skipped, so a broad send can never produce an unexpected bill. A campaign that wants to mail a million strangers at an average price of five cents sees a $50,000 total before it commits. Mass distribution to people who never asked for it then costs something in proportion to the attention it takes.

The Treasury collects nothing. Postage moves from sender to recipient, and providers may keep no part of it. That makes attention postage something other than an email tax, a phrase with its own history: for years a hoax about a fictitious "Bill 602P" claimed that Congress would surcharge every email, and the Postal Service had to deny it publicly (Chief Marketer).

What the research says

The idea that a sender should pay to reach an unwilling reader is older than most of the spam it was meant to stop. In 1992 Cynthia Dwork and Moni Naor proposed that a sender's computer solve a moderately hard puzzle before a message is accepted, a computational postage that became the ancestor of proof-of-work (Dwork and Naor, CRYPTO '92). Scott Fahlman proposed the version closest to this one in 2002. "My attention is a valuable commodity," he wrote, and each recipient should set the price a sender must offer, declining the payment when glad to hear from the sender (Fahlman, IBM Systems Journal, 2002). Thede Loder, Marshall Van Alstyne and Rick Wash formalized the idea as an attention bond and showed that "in certain limited cases" a recipient-set price produces greater welfare "than use of even 'perfect' filters" (Loder, Van Alstyne and Wash, 2006).

The economics explain why a small price is enough. Justin Rao and David Reiley estimated that spam cost American firms and consumers almost $20 billion a year while bringing spammers and their merchants on the order of $200 million worldwide, an external cost roughly 100 times the private gain (Rao and Reiley, Journal of Economic Perspectives, 2012). Chris Kanich, Stefan Savage and colleagues infiltrated a working botnet and measured its yield directly: "out of 350 million pharmacy campaign e-mails only 28 conversions resulted" (Kanich et al., ACM CCS, 2008). At one cent per message that campaign would have cost $3.5 million, or $125,000 per sale (derived: 350 million × $0.01 ÷ 28).

In a laboratory experiment, Robert Kraut and colleagues found that "charging postage for e-mail causes senders to be more selective and to send fewer messages." They also found that recipients did not read postage as a sign that a message was important (Kraut et al., Human-Computer Interaction, 2005). The effect works through the sender's decision, which is why the price must be visible before the sender commits.

Two failures shaped this design. Ben Laurie and Richard Clayton showed in 2004 that computational postage cannot do the job alone: a puzzle costly enough to deter spammers, who command hijacked computers by the million, also burdens legitimate senders unless they are excused from it (Laurie and Clayton, 2004). Money with a free path for known correspondents avoids that trap, because ordinary mail never pays. Goodmail's CertifiedEmail, which AOL adopted in 2006 and Yahoo later joined, charged senders a fraction of a cent per message to bypass spam filtering and shared the fee with the mailbox providers. The Electronic Frontier Foundation argued that a provider paid per certified message had a reason to let its free filtering decline (eWeek). Yahoo ended its deal in early 2010 and Goodmail later shut down (netimperative). Paying the recipient, and barring the provider from any share, removes that conflict of interest.

Per-message pricing already runs at scale. Since 2014 LinkedIn has returned a sender's InMail credit when the recipient responds within 90 days (LinkedIn, 2014). U.S. wireless carriers charge registered business texters a fraction of a cent per message (8x8). Neither system pays the person whose attention is taken.

The research carries one more lesson, which is that fraud moves to whichever channel is cheapest. Reported losses to scams that began on social media reached $2.1 billion in 2025, about eight times the 2020 level (FTC, April 2026). Losses to text-message scams reached $470 million in 2024, more than five times the 2020 figure (FTC, April 2025). A price on one channel pushes fraud onto the next, which is the argument for attaching the right to the person and applying it on every channel where a stranger can send a message directly.

Why now

Fred Heiding, Bruce Schneier and colleagues tested fully automated, AI-written spear phishing in a study of 101 participants. The AI-written emails drew a 54 percent click-through rate, the same as emails written by human experts and four and a half times the 12 percent rate for the control group. They estimate that AI raises the profitability of large phishing campaigns by up to 50 times (Heiding et al., Expert Systems with Applications, 2026). Researchers at Columbia, the University of Chicago and Barracuda Networks found that 51 percent of spam in April 2025 was AI-generated (Columbia Engineering, 2025). AI pushes the cost of a convincing, personal message toward zero, and attention postage puts a floor under exactly that cost. Because each person sets the price, people can raise it as AI makes each message more valuable to a fraudster. Filters stay in place to catch malware and impersonation in paid mail, and they no longer carry the whole defense.

The scale of the problem is now measured in hundreds of billions of dollars. Bloomberg reported on October 10, 2026, that victims in 42 countries lost $442 billion to scams last year, more than some valuations of the global cocaine trade. The figure is the Global Anti-Scam Alliance's estimate from consumer surveys (PYMNTS, summarizing Bloomberg). In the United States, consumers reported $15.9 billion in fraud losses to the FTC in 2025, and the FTC estimates that the full cost of fraud in 2024, counting losses never reported, could be as high as $195.9 billion (FTC testimony to the Joint Economic Committee, March 2026).

The infrastructure already exists. Gmail and Yahoo have required bulk senders to authenticate their domains since February 2024, so a receiving server can already confirm which domain a message really comes from (dmarcian). The GENIUS Act, signed in July 2025, set federal rules for licensed issuers of payment stablecoins (Debevoise). The x402 standard that Coinbase published in May 2025 lets software pay per request in stablecoins, with on-chain fees near one cent on networks such as Base (Coinbase). The IETF's Privacy Pass architecture describes tokens that prove a payment was made without revealing who made it (RFC 9576). AI agents that send email on their owners' behalf are arriving in the same years. Under attention postage, an agent that emails strangers spends its owner's money, inside a budget its owner sets.

What it would stop, and what it would not

For bulk email from strangers, a reduction of ninety percent or more is a reasonable expectation. Kaspersky counted spam at 44.99 percent of global email traffic in 2025 (Kaspersky, 2026). Most of that is low-yield bulk mail of the kind Kanich measured, which cannot survive a price measured in cents. The share that would remain is mail from senders who expect a single message to be worth more than its price.

Fraud losses are a different measure, and the reduction there is much smaller, because most losses now begin outside email.

Sources for the table: the FTC's Consumer Sentinel Data Book 2024 and the FBI's 2025 Internet Crime Report. An email-only rule would therefore reduce reported U.S. fraud losses by a few percent, and by more only to the extent that business email compromise arrives from lookalike domains rather than hijacked accounts (derived from the table). That is real money, and it is far short of ninety percent.

The reach grows when the same right applies to calls, texts and direct messages from strangers. In 2025 FTC reports of money lost to scams, the scam began with a phone call in 11 percent of cases, an email in 10 percent and a text in 7 percent (FTC, April 2026). Those add to 28 percent, before counting direct messages on social media, which account for part of another 28 percent. A rule covering calls, texts, email and direct messages would reach the first contact in somewhere between 28 and 56 percent of loss reports (derived). How much of that reach becomes prevented loss depends on the prices people set compared with what they are worth to a fraudster. No deployment exists yet to measure it, and part of what is blocked will move to advertising and fake websites. Those channels are the subject of the Accord's separate work on algorithmic harm.

Targeted fraud is the limit. The average business email compromise complaint in 2025 was about $123,000 (derived: $3,046,598,558 ÷ 24,768 complaints), and a fraudster expecting that return will pay $5 to reach a chief financial officer. Postage still works against him. He pays for every message to the executives who never respond. His message arrives labeled as a paid first contact, which is hard to pass off as a note from a colleague. If his token was bought from an identified account, it leaves a payment trail at the first contact. Kirill Levchenko and colleagues found that 95 percent of spam-advertised pharmaceutical, replica and software products were monetized through merchant services at a small number of banks, which identified payment as the narrowest point in the spam business (Levchenko et al., IEEE Symposium on Security and Privacy, 2011).

Technical requirements

The statute imposes two duties on every email provider above a size threshold. Receiving servers refuse mail from strangers that lacks valid postage, and sending services attach postage when the sender pays it. Everything else exists to make those duties work.

Receiving side

• Keep a free list for each account, covering contacts, correspondents, subscriptions and permission addresses, with a per-sender switch that returns any listed sender to paying.
• Answer price queries for mail from strangers. A query returns a price whether or not the address exists, so the lookup cannot be used to harvest valid addresses.
• Refuse unpaid mail from strangers at the server, with a notice that states the price and how to pay, so a legitimate sender learns the rule and does not silently lose the message.
• Verify each token: signed by a licensed issuer, bound to this message and this recipient, at or above the recipient's price, and never spent before.
• Credit the recipient, show the message as a paid first contact, and apply the recipient's refund setting.

Sending side

• Look up prices as addresses are entered and show the total before the sender presses send.
• Enforce the sender's maximum per recipient and a daily spending cap, so a hijacked account cannot drain its owner's balance.
• Buy tokens from a licensed issuer and attach one to each message bound for a stranger.

Payment layer

• Tokens are prepaid, so a payment cannot be reversed after delivery.
• Issuers settle net balances with one another and with providers in batches.
• Anonymous tokens are allowed, and a recipient may set a separate price for them. Anonymous speech stays available, and each person decides what it costs to reach them anonymously.

Smart contracts and stablecoins suit the settlement layer. They cannot carry each message. The world sent and received about 376 billion emails a day in 2025 (Statista), more than 4 million a second. The entire Ethereum ecosystem, main chain and layer-2 networks together, processed about 4,236 transactions a second on October 12, 2026 (growthepie). If only one message in ten needed postage, settling each on-chain would take about 100 times that capacity (derived). Tokens are therefore signed off-chain, checked by the receiving server in milliseconds, and settled in batches on whatever rail the issuer uses: a bank, a card network or a licensed stablecoin.

Where it sits in the Accord

Attention postage extends the Externality Limiter, the Accord's engine for pricing harms where they originate, into the cognitive domain. Oversight belongs to the Digital Safety Board, the Accord's independent expert panel for digital harms. The Board has authority to certify the technical standard for price queries, tokens and settlement, to audit covered providers for conformance, and to publish its methodology. It sets no one's price and makes no judgment about content. Its determinations are evidence-based, methodology-audited and appealable. The FTC, which enforces the CAN-SPAM Act today, enforces the provider duties.

In 2003 Congress chose, in the CAN-SPAM Act, an opt-out approach "in which commercial email is permitted until a recipient acts to stop it," and preempted California's stricter opt-in requirement (WilmerHale, 2003). That choice left the cost of unwanted mail with the people receiving it for more than two decades. Attention postage returns the decision to them.

Sources

• Dwork, C. and Naor, M. (1992). "Pricing via Processing or Combatting Junk Mail." Advances in Cryptology, CRYPTO '92, pp. 139–147. DOI 10.1007/3-540-48071-4_10
• Fahlman, S. E. (2002). "Selling interrupt rights: A way to control unwanted e-mail and telephone calls." IBM Systems Journal 41(4): 759–766. DOI 10.1147/sj.414.0759
• Heiding, F., Lermen, S., Kao, A., Verdun, C. M., Schneier, B. and Vishwanath, A. (2026). "Evaluating Large Language Models' Ability to Automate Spear Phishing." Expert Systems with Applications 314: 131546. Author archive
• Kanich, C., Kreibich, C., Levchenko, K., Enright, B., Voelker, G. M., Paxson, V. and Savage, S. (2008). "Spamalytics: An Empirical Analysis of Spam Marketing Conversion." ACM CCS 2008. DOI 10.1145/1455770.1455774
• Kraut, R. E., Sunder, S., Telang, R. and Morris, J. (2005). "Pricing Electronic Mail to Solve the Problem of Spam." Human-Computer Interaction 20: 195–223. PDF
• Laurie, B. and Clayton, R. (2004). "Proof-of-Work Proves Not to Work." Workshop on the Economics of Information Security. PDF
• Levchenko, K. et al. (2011). "Click Trajectories: End-to-End Analysis of the Spam Value Chain." IEEE Symposium on Security and Privacy. DOI 10.1109/SP.2011.24
• Loder, T., Van Alstyne, M. and Wash, R. (2006). "An Economic Response to Unsolicited Communication." The B.E. Journal of Economic Analysis and Policy 6(1). DOI 10.2202/1538-0637.1322
• Rao, J. M. and Reiley, D. H. (2012). "The Economics of Spam." Journal of Economic Perspectives 26(3): 87–110. DOI 10.1257/jep.26.3.87
• Rowan v. United States Post Office Department, 397 U.S. 728 (1970). Opinion
• Federal Trade Commission, Consumer Sentinel Network Data Book 2024 (March 2025); Data Spotlights of April 2025 and April 2026; testimony before the Joint Economic Committee, March 25, 2026.
• Federal Bureau of Investigation, Internet Crime Complaint Center, 2025 Internet Crime Report.

Figures marked derived are calculated in this post from the cited sources and are not estimates made by those sources. Reported fraud losses undercount actual losses; the FTC estimates the full 2024 cost could be as high as $195.9 billion against $12.5 billion reported. The $442 billion global figure comes from consumer surveys in 42 countries and is not comparable with the FTC and FBI complaint data. The reach figures for a rule covering every channel are bounds on where postage applies, not forecasts of prevented loss, because no attention-postage system has yet operated at scale.