---
title: "Plain-text passwords: lessons from a €5M CNIL fine"
description: "A €5M CNIL fine for plain-text passwords, shared logins, and data kept too long. The questions to ask your developer about your homegrown business apps."
canonical_url: "https://www.nexperio.com/en/blog/mots-de-passe-clair-cnil-pme"
language: en-US
author: Arnaud Guedj
date_published: 2026-11-20T08:15:02.324+00:00
date_modified: 2026-11-20T08:15:02.759052+00:00
---

# Plain-text passwords: what France's €5M fine teaches small businesses

Storing customer passwords in plain text, sharing one login among several employees, and keeping data with no time limit: that is exactly what France's data protection authority, the CNIL, just fined €5 million, in a decision issued on November 12, 2026 ([Siècle Digital](https://siecledigital.fr/2026/11/18/la-macif-ecope-de-5-millions-deuros-damende-pour-avoir-stocke-des-mots-de-passe-en-clair/)) and [published on November 18, 2026](https://www.cnil.fr/fr/conservation-des-mots-de-passe-sanction-5-millions-deuros). You don't need to be a large insurer to be affected. If your small business runs a customer portal, an extranet, or any in-house tool, you can check these three points with your developer by asking a few plain questions.

The company involved is Macif Vie, which confirmed it was the company concerned and says no customer data was disclosed ([franceinfo, November 18, 2026](https://www.franceinfo.fr/internet/securite-sur-internet/cyberattaques/la-cnil-inflige-une-amende-de-5-millions-d-euros-a-l-assureur-macif-vie-en-raison-de-l-insuffisance-des-mesures-prises-pour-la-securite-des-donnees_8288285.html)). But this post is about you. The CNIL published its decision for educational purposes, so let's take it at its word and run a checkup on your homegrown apps, whether they came from an agency, a freelancer, or an AI assistant.

## What exactly did the CNIL hold against the sanctioned company?

Three things: customer passwords stored in plain text, logins shared among employees to access customer databases, and data kept far too long. The legal basis is GDPR Articles 32 (security) and 5(1)(e) (storage limitation).

After being alerted to how customer passwords and the answers to account security questions were stored, the CNIL ran an inspection ([press release, November 18, 2026](https://www.cnil.fr/fr/conservation-des-mots-de-passe-sanction-5-millions-deuros)). Its restricted committee, the body that issues penalties, found that the passwords used to access customer accounts were stored in plain text. [Siècle Digital](https://siecledigital.fr/2026/11/18/la-macif-ecope-de-5-millions-deuros-damende-pour-avoir-stocke-des-mots-de-passe-en-clair/) reports nearly 1.8 million passwords stored without any hashing function.

The company also kept contact form data for more than 10 years, and kept login data beyond the one-year limit it had set for itself. It fixed things during the proceedings, including a monthly purge of customer data.

## What is a "plain-text" password, and how should it be stored?

A plain-text password is saved exactly as the user typed it, readable by anyone who opens the database. The CNIL's rule: it must be hashed, with a salt, using a dedicated function.

**Hashing** is a one-way transformation: the password becomes a fingerprint that can't be turned back into the original. At each login, the site computes the fingerprint of what you type and compares it with the stored one. It never needs to know your actual password.

A **salt** is a random value, unique to each user, added to the password before hashing. It stops attackers from using the precomputed fingerprint tables that circulate online.

The [CNIL recommendation of July 21, 2022 (decision no. 2022-100)](https://www.cnil.fr/sites/default/files/atoms/files/deliberation-2022-100-du-21-juillet-2022_recommandation-aux-mots-de-passe.pdf) is blunt: a password must never be stored in plain text by the organization handling it. It calls for a secure, non-reversible function, a salt of at least 128 bits per user, and time or memory cost settings that should be reviewed regularly.

The [CNIL guide on encryption, hashing, and signatures](https://www.cnil.fr/fr/securite-chiffrement-hachage-signature) recommends bcrypt, scrypt, Argon2, or PBKDF2. It rules out MD5 and SHA-1, which are obsolete, hashing without a salt, and confusing hashing with encryption, which can be reversed with a key.

A simple test: if your tool can email you your current password, that's a symptom. A well-built system doesn't know it.

## Why is a login shared among coworkers a problem?

Because nobody can tell who did what. When an incident happens, the investigation hits a wall at a username five people were using.

The CNIL puts it plainly: only individual accounts allow effective traceability of access and actions on a system, especially when investigating a security incident or data breach.

In small businesses, the shared "admin" account on the back office is a classic. It's convenient until an employee leaves: the password should change for everyone, and nobody does it. Named accounts, each with limited permissions, solve both problems.

## How long should you keep contact form and login data?

As long as you need it, and no longer. For a contact form, the CNIL considers keeping it more than 10 years excessive once the customer's request has been handled.

There's no universal retention period to copy and paste. Once data is no longer needed, you anonymize it, archive it if a legal obligation requires it, or delete it.

The second lesson stings a little more: the company wasn't following its own one-year rule. **Writing a retention period into your privacy policy isn't enough: your tool has to actually run the purge.** In many homegrown apps, nobody ever coded that deletion, and data quietly piles up. The same gap between policy and technical reality showed up in [the CNIL fine against Fnac Darty over cookies](https://www.nexperio.com/blog/cookies-recommandations-produits-cnil-ecommerce).

## Why can't these checks wait anymore?

Because the window between a security fix and its exploitation is shrinking. AI tools can now automate the whole chain, with no human in the loop.

On November 16, 2026, the Google Threat Intelligence Group (GTIG), working with Mandiant, published [a report](https://cloud.google.com/blog/topics/threat-intelligence/agentic-threat-activity-crosses-a-significant-threshold/) describing a first. On an exposed server, researchers found a set of AI agents called "exploit_agent." It continuously compares the code of popular web software before and after a **patch** is released (a patch is the update that closes a security hole), spots the fixed flaw, writes the attack, and runs it.

Flaws that have already been fixed are called "n-day" vulnerabilities: they're known and repaired, yet still open wherever the update hasn't been installed. For attackers, they pay off better than unknown flaws, because the work boils down to comparing old code, new code, and release notes. GTIG says these tools could significantly reduce the time between a patch and its exploitation. According to [Siècle Digital (November 18, 2026)](https://siecledigital.fr/2026/11/18/une-ia-capable-de-mener-seule-des-cyberattaques-frappe-des-sites-francais-selon-google/), the operator targeted French websites in media, entertainment, and education.

So security updates get installed right away, not "next sprint." I covered this for PHP sites in [my post on PHP vulnerabilities](https://www.nexperio.com/blog/failles-php-site-ecommerce-pme).

## What should you ask your developer?

Four jargon-free questions are enough for a first diagnosis.

| Question to ask | Good answer | Answer that should worry you |
|---|---|---|
| How are user passwords stored? | "Hashed with bcrypt, scrypt, Argon2, or PBKDF2, with a salt per user." | "Encrypted," "MD5," or an awkward silence |
| Who logs into the database, with which accounts? | "One account per person, with appropriate permissions." | "There's a shared admin account" |
| How long do we keep the data? | "This long, with an automatic purge." | "We keep everything, just in case" |
| When do you install security patches? | "As soon as they're released, following a set procedure." | "When we get around to it" |

## My prescription

1. **Send these four questions to your developer in writing**, and ask for written answers.
2. **Eliminate shared accounts** and create one account per person. While you're at it, close access for former employees and contractors.
3. **Have the purge coded** to match the retention period stated in your privacy policy, then check that it runs.
4. **Schedule security updates** so they're installed promptly.
5. **Read the CNIL texts** linked above. I'm not a lawyer: they are the reference.

These are exactly the questions any health check of a homegrown app starts with, covering security, backups, GDPR, and hosting. The good news: the remedies are well known and well documented.

## Frequently asked questions

### How can I tell if my website stores passwords in plain text?

Ask your developer in writing. The right answer: passwords are hashed with bcrypt, scrypt, Argon2, or PBKDF2, with a unique salt for each user, as France's CNIL recommends. Be wary of answers like "encrypted" or "MD5." Another clue: if your tool can email you your current password, it knows that password, which means it stores it in a reversible way.

### Why is an admin account shared by several employees a problem?

Because it makes it impossible to know who did what. The CNIL points out that only individual accounts provide effective traceability of access and actions, especially when investigating a security incident. A shared account also complicates departures, since the password would have to change for everyone. One named account per person, with limited permissions, solves both problems.

### How long can I keep messages from my contact form?

As long as you need to handle the request, and no longer. In its November 2026 penalty, the CNIL found that keeping this data for more than 10 years after the request was handled was excessive. After that, you anonymize it, archive it if a legal obligation requires it, or delete it. Above all, the retention period in your privacy policy must be enforced by an actual purge in the tool.

### Why should security updates be installed right away?

Because the window between a patch release and its exploitation is shrinking. On November 16, 2026, Google described an AI tool that compares code before and after a patch, spots the fixed flaw, writes the attack, and runs it with no human involved. A flaw that has been fixed upstream but not yet installed on your systems becomes a prime target for this kind of tool.

## Structured data (JSON-LD)

```json
{
  "@context": "https://schema.org",
  "@type": "BlogPosting",
  "@id": "https://www.nexperio.com/en/blog/mots-de-passe-clair-cnil-pme#article",
  "mainEntityOfPage": {
    "@type": "WebPage",
    "@id": "https://www.nexperio.com/en/blog/mots-de-passe-clair-cnil-pme"
  },
  "url": "https://www.nexperio.com/en/blog/mots-de-passe-clair-cnil-pme",
  "headline": "Plain-text passwords: lessons from a €5M CNIL fine",
  "description": "A €5M CNIL fine for plain-text passwords, shared logins, and data kept too long. The questions to ask your developer about your homegrown business apps.",
  "datePublished": "2026-11-20T08:15:02.324+00:00",
  "dateModified": "2026-11-20T08:15:02.759052+00:00",
  "inLanguage": "en-US",
  "author": {
    "@type": "Person",
    "name": "Arnaud Guedj",
    "url": "https://www.nexperio.com/en/le-docteur-digital"
  },
  "publisher": {
    "@type": "Organization",
    "name": "Nexperio",
    "url": "https://www.nexperio.com",
    "logo": {
      "@type": "ImageObject",
      "url": "https://www.nexperio.com/assets/nexperio-logo.png"
    }
  },
  "image": {
    "@type": "ImageObject",
    "url": "https://qarqnjzixkkunhdzgule.supabase.co/storage/v1/object/public/blog/mots-de-passe-clair-cnil-pme/cover-6684c7bc2190.jpg",
    "contentUrl": "https://qarqnjzixkkunhdzgule.supabase.co/storage/v1/object/public/blog/mots-de-passe-clair-cnil-pme/cover-6684c7bc2190.jpg",
    "caption": "In a bright workshop office in morning light, a business owner places individual keys into a wall of small oak lockers, each with its own lock, instead of a single shared key ring left hanging on a hook."
  }
}
```

```json
{
  "@context": "https://schema.org",
  "@type": "FAQPage",
  "mainEntity": [
    {
      "@type": "Question",
      "name": "How can I tell if my website stores passwords in plain text?",
      "acceptedAnswer": {
        "@type": "Answer",
        "text": "Ask your developer in writing. The right answer: passwords are hashed with bcrypt, scrypt, Argon2, or PBKDF2, with a unique salt for each user, as France's CNIL recommends. Be wary of answers like \"encrypted\" or \"MD5.\" Another clue: if your tool can email you your current password, it knows that password, which means it stores it in a reversible way."
      }
    },
    {
      "@type": "Question",
      "name": "Why is an admin account shared by several employees a problem?",
      "acceptedAnswer": {
        "@type": "Answer",
        "text": "Because it makes it impossible to know who did what. The CNIL points out that only individual accounts provide effective traceability of access and actions, especially when investigating a security incident. A shared account also complicates departures, since the password would have to change for everyone. One named account per person, with limited permissions, solves both problems."
      }
    },
    {
      "@type": "Question",
      "name": "How long can I keep messages from my contact form?",
      "acceptedAnswer": {
        "@type": "Answer",
        "text": "As long as you need to handle the request, and no longer. In its November 2026 penalty, the CNIL found that keeping this data for more than 10 years after the request was handled was excessive. After that, you anonymize it, archive it if a legal obligation requires it, or delete it. Above all, the retention period in your privacy policy must be enforced by an actual purge in the tool."
      }
    },
    {
      "@type": "Question",
      "name": "Why should security updates be installed right away?",
      "acceptedAnswer": {
        "@type": "Answer",
        "text": "Because the window between a patch release and its exploitation is shrinking. On November 16, 2026, Google described an AI tool that compares code before and after a patch, spots the fixed flaw, writes the attack, and runs it with no human involved. A flaw that has been fixed upstream but not yet installed on your systems becomes a prime target for this kind of tool."
      }
    }
  ]
}
```
