The Internet Forgot How to Forget

Spent convictions, credit reports that purge themselves, gossip that died with its generation — forgetting used to be engineered into our institutions on purpose. Nobody abolished it. The internet just never built it.

Somewhere in Britain this morning, somebody is filling in a job application. They reach the question — have you ever been convicted of a criminal offence? — and they tick “no,” and it is not a lie, even though they were. Under the Rehabilitation of Offenders Act 1974, most convictions expire after a set number of years: they become “spent,” and from then on, with some exceptions for sensitive work, the law lets a person treat them as never having happened. Parliament did not misplace these records. It scheduled their forgetting.

We rarely notice how much engineering went into forgetting, because it looks like nothing — like an absence. But the machinery is everywhere, once you look. Since 1970, America’s Fair Credit Reporting Act has barred credit bureaus from mentioning most adverse items older than seven years, and bankruptcies older than ten; Congress wrote decay directly into the database. (Criminal convictions, tellingly, are exempt — America forgets your debts but not your crimes. Britain, with its spent convictions, made roughly the opposite choice.) Juvenile records get sealed. Bankruptcy discharge exists to make obligations stop existing. Gossip died with the generation that traded in it, and moving two counties over functioned as a reset. In a 2002 paper, “The Social Benefits of Forgetfulness,” Jean-François Blanchette and Deborah Johnson catalogued all of this as what it was: technology for letting people begin again. Forgetting was load-bearing.

None of that happened by accident — which is what makes the internet’s arrangement so strange. For nearly all of history, remembering was expensive and forgetting was free. Parchment rotted, scribes billed by the page, archives flooded and burned; keeping a record meant deciding to. Digital storage inverted the economics completely. Retention now costs effectively nothing — it is simply what happens when nobody does anything — while deletion is work: someone must decide, build the process, chase the copies. As one paper on web memory puts it, “remembering has become the norm”; the web itself “does not have any procedure to forget.” No legislature repealed forgetting. No company announced the end of the fresh start. The default flipped, and the flip was nobody’s decision — which is precisely the problem.

A hoarder with amnesia

Here is the complication that should unsettle the tidy version of this story, the one where the internet remembers everything. It doesn’t. It forgets constantly, voraciously, and at random. In a 2013 paper, the legal scholar Meg Leta Ambrose reported one persistence study in which 85 percent of online content disappeared within a year — 59 percent within a week. Link rot hollows out court citations and news archives; platforms fold and take a decade of photographs with them; the forum thread, the band page, the first draft of your life online — gone, unmourned. Meanwhile the mugshot is forever, because somebody monetizes it, and the screenshot outlives the apology. What survives is not what matters but what somebody profits from keeping. The old forgetting was patterned: the spent conviction faded on schedule while the land registry endured. The new arrangement manages the worst of both — memory and amnesia, each allocated by accident. The internet is not an elephant. It is a hoarder with amnesia.

The right to be harder to find

Europe has spent a decade trying to retrofit intention onto this default. The landmark is Google Spain, decided by the EU’s Court of Justice in 2014: Mario Costeja González objected that searching his name surfaced a 1998 newspaper notice about a long-settled debt, and the court agreed he could have the link delisted. The remedy is widely misdescribed — the newspaper’s pages stayed in the archive; what died was their findability under his name. The GDPR later codified a fuller right to erasure — Article 17, the “right to be forgotten” — for data no longer necessary, unlawfully processed or held past consent, with carve-outs for expression, history and statistics, plus a duty to take “reasonable steps” to chase copies once data has gone public. In 2019, the same court ruled that delisting stops at the EU’s borders. Forgetting, it turns out, is territorial. The internet is not.

And the hard cases are where erasure meets the record itself. In a case from Belgium, courts ordered a newspaper to anonymize an old archived article about a person’s past crime; the European Court of Human Rights upheld the order, finding no violation of press freedom and accepting that an unaltered archive can function as a “virtual criminal record.” The judges weighed truth, public interest, historical value, harm and elapsed time — which is to say they did the thing institutions used to do: they decided. Free-press instincts should still twitch; a court telling a newsroom to alter its archive is not nothing. American law mostly flinches from that — expungement there seals the state’s files but leaves the news archive alone, and the credit-report limits bind bureaus, not reporters. Each jurisdiction draws the line somewhere different. The point is that there is a line, and someone draws it on purpose.

There is a clarifying way to see what all this law is straining toward. Paper-era forgetting rarely meant destruction. The courthouse file existed; you just had to drive to the courthouse. Lawyers call the old condition “practical obscurity” — available in principle, invisible in practice — and it was the default state of nearly everything ever recorded about anyone. Search ended that. Everything is now one query away from everyone, permanently, and true deletion is genuinely hard once data escapes: backups, caches, replicas, screenshots. Article 17’s “reasonable steps” is an honest admission that public data can be pursued but never quite recalled. The right to be forgotten is, in practice, the right to be harder to find — an attempt to rebuild practical obscurity with court orders, now that it no longer comes free with the physics of paper.

Which brings things back to the applicant and the box. The law spent decades calibrating when they may honestly answer “no” — how many years must pass, which offenses, which jobs are exempt — because reintegration was considered worth a statute. The search engine calibrates nothing. It answers in a quarter of a second, forever. And forgetting will happen to the internet regardless: links rot, platforms die, drives fail. The only live question is whether it happens on a schedule, for reasons, in somebody’s favor — or at random, to whoever’s luck runs out. Nobody ever voted for the permanent record. Permanence wasn’t a policy; it was a price point. A default is just a decision nobody admitted to making.