INTRODUCTION
“At least one uppercase letter, one number, one symbol, no spaces, and not one of your last five passwords.” Every product team that writes a rule like this means well. The rule looks like security. In practice it trains people to type Summer2026!, write it on a sticky note, and change it to Autumn2026! when the form forces them to.
The standard most of these rules were once copied from now says the opposite. NIST’s Special Publication 800-63B-4, the US federal guideline on authentication finalised in August 2025, is blunt: services “SHALL NOT impose other composition rules” and “SHALL NOT require subscribers to change passwords periodically”. If your sign-up form still has a checklist of character types, it is enforcing advice that its own source has withdrawn.
This piece is for founders and product leads who own a sign-up or account settings flow. The decision is simple to state: which password rules should stay, which should go, and what should replace them.
Why the old rules backfire
Composition rules assume that people pick characters at random. They do not. Asked for an uppercase letter, most people capitalise the first one. Asked for a number and a symbol, they add “1!” at the end. The result satisfies the checklist and stays predictable, because attackers guess the same patterns people use.
Forced rotation has the same problem over time. A password that must change every 90 days becomes a password with a counter in it. The new one is only as strong as the old one plus a guessable increment, and the cost lands on support: more resets, more lockouts, more “I can’t log in” tickets.
Both rules make the product feel strict while making the actual secret weaker.
What to require instead
NIST replaces the checklist with four mechanisms that do real work.
Length. At least 15 characters when the password is the only factor; at least 8 when it is used with a second factor. Allow at least 64, and accept spaces and Unicode, so long passphrases work.
A blocklist. Check new passwords against known breached passwords, dictionary words and words specific to your product, such as its name or the person’s username. If one matches, say why it was refused.
Rate limiting. Limit failed attempts per account. A blocklist only has to stop guesses an attacker could reach before that limit.
Change on evidence. Force a new password when there is a sign of compromise, not on a calendar.
GOV.UK’s design system lands in the same place for its services: set a minimum length, do not set a maximum, block common passwords, explain your restrictions and do not make people keep changing their passwords. Its suggested minimum is 8 rather than 15; NIST’s higher bar applies when nothing else protects the account.
Let the password manager do its job
The strongest password most people will ever use is the one a password manager generates and fills in. NIST now requires services to allow password managers and autofill, recommends allowing paste, and recommends offering a way to show the password while typing.
In practice that means removing the script that blocks paste, removing the “confirm password” field if it exists only to catch typos a show-password toggle catches better, and marking fields with autocomplete="new-password" on sign-up and autocomplete="current-password" on sign-in, as described in MDN’s reference for the autocomplete attribute. Those two values tell the browser when to offer a generated password and when to fill a saved one.
Delete the hints and the questions
Password hints and “What was the name of your first pet?” look like recovery features. They are second passwords with less protection. NIST forbids hints readable by anyone who is not signed in and forbids prompting people to set security questions. GOV.UK says the same in plainer terms: the answers are too obscure to remember or too easy for someone else to find.
A time-limited reset link or code sent to a verified address, plus a message telling the person a reset happened, covers the same need. Recovery deserves more design attention than it usually gets; we made that argument for passkeys in Passkeys still need a way back in, and it applies here too.
The serious counterargument
Some customers will ask for the old rules. Enterprise security questionnaires, older internal policies and some compliance frameworks still list complexity and rotation, and a sales team does not want to argue with a procurement checklist. NIST’s guideline is written for US federal systems; it is not automatically the standard your customer is audited against.
That is a real constraint, and the answer is not to ignore it. Find out what the framework your customer uses actually requires today, apply that requirement where it applies, and keep it out of everyone else’s sign-up flow. Often the stronger answer to “we need complexity rules” is a second factor or passkeys; GOV.UK explicitly suggests adding a second factor rather than extra password restrictions.
The other objection is that 15 characters feels long. It does, if people type it. With a password manager or a passphrase of four ordinary words, it stops being a burden; with a second factor, 8 is allowed.
A sign-up form, before and after
Hypothetical example. Before: an 8-to-16-character password with four character classes, no paste, a confirmation field, a hint, and expiry every 90 days. After: a single field marked new-password, minimum 15 characters, maximum 128, spaces allowed, a show-password toggle, a blocklist check that says “This password appears in known data breaches; choose another”, and an optional second factor offered right after sign-up. The form has fewer rules and a harder secret.
What to do this week
Open your own sign-up and password-change screens and check five things: Is there a composition rule? Is there forced rotation? Is paste blocked? Is there a hint or security question? Is there a blocklist? Every “yes” to the first four and every “no” to the last is a rule that feels secure and is not. Remove them, and write down the one compliance requirement, if any, that genuinely forces you to keep one.
Sources checked on 17 November 2026: NIST SP 800-63B-4 (pages.nist.gov, August 2025), the GOV.UK Design System passwords pattern and MDN’s autocomplete reference.