Accept All and Report for Duty

A student clicking a mouse is handed a Turkish army uniform and boots through the computer screen.

By Martyn Walker

I have two ways of finding out what time the chemist shuts. On the door it says half past five. On the website it says, ‘We value your privacy.’ I have always rather liked the door.

When Britain was working out how to apply the EU’s cookie directive, Downing Street saw an early draft of the proposed rules and, I gather, thought them bonkers. I should love to know how the uncivil service minuted that. ‘Ministers welcomed the opportunity for further reflection’, at a guess. A few tiers of advisers later, I was asked to write a white paper on what the proposals would mean for businesses trading online.

A cookie is a small piece of data your browser keeps. Some last for a visit. Others are there when you come back tomorrow. The useful one in an online shop often amounts to a cloakroom ticket. The shop keeps the basket and your browser presents the ticket, so the shoes you put in it are still there when you go back for the polish. I’m quite content with that arrangement. I might even ask to be told when Manolo Blahnik produces another red abomination, if only to see what it costs. I draw the line at the bloody thing turning up beside the cricket scores because I once looked at it in a shop window. [1] [2]

The EU recognised the difference. Its 2009 directive required consent to store or read information on a device, with an express exception for what was strictly necessary to provide a service the visitor had asked for. My shopping basket qualified. Brussels left the design of the notice open. Whoever decided that reading the cricket required a conference with the advertising industry exercised considerable initiative. [3] [4]

I had a practical question for my white paper. Would anybody actually read the thing? I had some willing students and we built a replica of the BBC News site using the BBC’s own stories. We told the students there were one or two errors in the material and asked them to record what they found. They had good reason to pay attention.

We made two changes. The cookie notice said that anyone accepting it agreed to serve a year in the Turkish army. Once they had accepted, we showed it to them again at random intervals of between ten and fifty pages. Even the most absent-minded recruit had a sporting chance of reconsidering.

The students found spelling mistakes. They took exception to the BBC’s grammar. Several stories were judged to be plain wrong, although on that point we had only the students’ word and the BBC was entitled to a second opinion. Every one of them clicked to accept. Nobody mentioned the Turkish army. I did briefly consider writing to Erdoğan with the good news about recruitment, but thought better of it.

On 24 May 2011, two days before the UK rules took effect, the government wrote to industry. Businesses working towards compliance would have time to get it right. A choice made through browser settings could count, provided the user actually chose them. Leaving the defaults alone would not do. I flatter myself that my paper helped. The department said it had settled its approach in April after consulting widely, so I shall resist the temptation to claim that I personally saved the British shopping basket. [5] [6] [7]

As a money laundering reporting officer, I have a fair amount of regulation to live with and good reason to value most of it. If a control produces an immaculate record of everybody saying yes, I want to know what happened to the people who wanted to say no. In a study published in 2020, researchers tried different consent screens on 40 people. Taking the refusal button off the first screen raised acceptance by 22–23 percentage points. You can see why a man who sells advertising might take an interest in where the buttons go. [8]

Did anyone lobby for the rule in the hope of selling the remedy? I have no evidence of that, though I recognise a good business when I see one. In 2025 the European Commission’s own staff estimated the annual cost to EU websites of complying with the cookie rule at €1.64 billion and the time people spent dealing with banners at 334 million hours a year. They allowed three and a half seconds per encounter. I should like to see the official who can read the list of ‘trusted partners’ in three and a half seconds. [9]

There is room for common sense. Under today’s UK rules a site can, for example, count visits solely to improve its service, subject to safeguards including clear information and a free, simple way to object. Advertising cookies still need consent, and the regulator expects refusal to be as easy as acceptance. A chemist who simply puts his opening hours online can leave me in peace. [4] [2] [10]

The trouble is that, after years of this nonsense, even I hesitate when the chemist’s website opens and tells me straight away what time he shuts. Is he trying to phish the merry fuck out of me? Give the fellow a handsome banner saying ‘We value your privacy’ and I feel better about him at once.

I really ought to know better. I wrote one of those notices myself and raised an army.

Sources

  1. How HTTP cookies work, MDN
  2. ICO guidance on cookies used in advertising
  3. Directive 2009/136/EC, Article 2(5) and recital 66
  4. ICO guidance on exemptions, including shopping baskets and service statistics
  5. UK government open letter to industry, 24 May 2011
  6. 2011 UK amending regulations, regulation 6
  7. UK government account of the enforcement approach
  8. Midas Nouwens and colleagues, Dark Patterns after the GDPR (2020)
  9. European Commission staff working document SWD(2025) 836, section on cookie banners and Table 6
  10. ICO guidance on offering accept and reject choices

Letter XXIII: The Login That Won’t Take No

A man stands at a rope barrier in front of an open doorway while a uniformed clerk offers him a second key.

By Martyn Walker
Published in Letters from a Nation in Decline

I look at my iPhone. It recognises me. Bitwarden fills in the long, unique password I chose never to have to remember. Then, between me and the account I came to use, up pops a banner inviting me to set up a passkey so that, in future, I can sign in more quickly.

I could have signed in more quickly this time. The banner was in the way.

I have been round this loop before. In August, when I changed my iPhone, back came the passkey settings I thought I had put to bed. The obvious switch also controlled the Bitwarden autofill I wanted to keep. By September I was asking, again, how to retain my passwords and two-factor authentication while dispensing with the invitations. By November I find myself wondering how a device capable of remembering my face has such difficulty remembering my answer.

There is a sound technical case for passkeys. A passkey is tied to the genuine website, so a convincing imitation cannot collect it and use it to sign in as me. That protection against phishing is real, even for someone who already uses strong passwords. It explains why the National Cyber Security Centre recommends passkeys. It does rather less to explain why I must keep declining one. Once I have made an informed choice, the next banner offers me no benefit whatever.

My father was a test pilot. One of his objections to bundling safety systems together was that each could make a perfectly good case for itself while the pilot had to deal with all of them at once. Eventually the distractions could make flying less safe. I think of him when a security prompt appears so often that my hand goes straight to the button that dismisses it. If you want me to read the next genuine warning, that is a peculiar habit to cultivate.

Then there is the question of whose security we are improving. I know where my Bitwarden passwords live and how I would recover them. If I accept a passkey on a shared tablet, who else can use it? GOV.UK’s own guidance warns that anyone able to unlock a shared device may also be able to use its passkey. I might quite reasonably let someone unlock a family tablet without intending to let them into my government account. In that household, accepting the invitation could weaken a boundary that mattered. I would like to weigh that decision at leisure, preferably somewhere other than halfway through doing something else.

The saving on the other side of the screen is easier to measure. In a FIDO Alliance case study, the insurer Branch reported that support tickets about authentication fell by half after it introduced passkeys. Its product manager identified fewer tickets as a principal business objective. Fair enough. Password resets cost money. What I should like to see alongside that figure is the total time spent by customers declining the offer. The help desk keeps accounts. Apparently my afternoon is free.

I would have liked regulators to intervene before government adopted the practice. In Britain we are already late. GOV.UK One Login offers passkeys, retains a password-and-code route and says explicitly that people who have yet to create a passkey will be prompted periodically when they sign in. A retailer that tries my patience may lose my custom. I have rather less scope to take my tax affairs elsewhere.

The Competition and Markets Authority should examine repeated enrolment prompts as part of its work on how online design steers consumer choices. The Government Digital Service can act on its own login screen now. Give “No, and don’t ask me again” the same prominence as “Set up a passkey”. Honour the answer across visits. Leave the option in account settings for anyone who changes their mind, and keep a secure alternative available to citizens who prefer it.

There is a petty tyranny in giving only one answer permanent effect. If I say yes, the asking stops. I should like the same courtesy when I say no.

Previous letter: Letter XXII: Provider, Provide — Do Not Prescribe.