The day you discover that a customer database has been copied, you are no longer dealing only with technology. You have to understand what was taken, check whether the intruder is still there, warn the right people and keep the business running. And customers may still receive fraud attempts months later.
The first hours are spent deciding
Questions arrive in no particular order, and almost all of them are urgent:
- Is the intruder still there, and should a service be shut down at the risk of stopping operations?
- Which data was viewed or copied, since when, and how many people are affected?
- What should we tell customers, partners and staff, and who answers the phone?
If there is a risk to people's rights and freedoms, the breach must be notified to the CNIL, France's data protection authority, without undue delay and, where feasible, within 72 hours. So decisions have to be made before all the answers are in.
The Cybermalveillance.gouv.fr guide for business leaders asks them to take into account an aspect that is often overlooked. A cyberattack can cause “an exceptional overload of activity and a feeling of shock, humiliation, incompetence, even guilt”. Decisions are made with incomplete information, and responsibility does not stop at 6 p.m.
The questions to settle before it happens are gathered in cyberattack, the questions to ask your teams before the crisis.
The people exposed live with the aftermath, often for longer
Cybermalveillance.gouv.fr, the French government's victim assistance platform, saw requests for help from individuals linked to data breaches rise by 107% in 2025. The data involved is often ordinary: name, address, phone number, sometimes bank details. It then feeds personalized phishing, fake bank adviser scams, payment fraud or identity theft.
The effects travel back to the company where the data came from. In a survey published by the CNIL in November 2025, 41% of respondents said they had already suffered fraudulent use of their data, and 57% of those who had suffered harm in the previous three years had given up a digital service.
AI increases the pace of fraud
Stolen data makes fraud more credible: a real name, a known supplier, sometimes a recent order. AI makes it easier to write personalized messages in large numbers. It does not invent this type of attack, but it can increase its pace.
ANSSI, France's national cybersecurity agency, observes that generative AI is already being used to profile victims, design social engineering content and develop malware. In February 2026, it also stated that no generative AI system, “official or jailbroken”, had carried out every stage of an attack on its own.
France's domestic intelligence agency, the DGSI, reports the case of an industrial site manager called on video by a fake executive, face and voice included, who demanded a funds transfer. The manager had the right reflex: end the call and check through the usual channels.
Your website has often become an information system
Many companies still talk about “their website”. Look at what it actually holds: years of order and invoice history, accounts of customers who left long ago, files uploaded through forms (contracts, ID documents, job applications), bank details, CSV exports generated for a one-off need and then forgotten on the server.
Each of these met a real need. Put together, they turn the platform into an information system exposed to the Internet. You do not need “sensitive data” in the GDPR sense (health, biometrics, political opinions) to build a fraud: a file combining name, email, phone number and purchase history is enough.
What a single access opens
Updates, multi-factor authentication, encryption, backups, monitoring: the basic measures make an intrusion less likely, or quicker to detect. None makes a system connected to the Internet invulnerable. The remaining question is what a compromised access can reach.
The sanction issued by the CNIL on September 3, 2026 against Hôpital privé de la Loire, a private hospital, gives a precise illustration. External access to the electronic patient record was protected by neither a VPN nor multi-factor authentication, and the rights of a single account opened the records of every patient. The attacker explored the system for several days. The data of 524,867 patients and 202,246 trusted contacts was compromised.
Among other things, the CNIL criticized an authorization policy that did not restrict access by care team.
Keeping less means exposing less
Deleted data can no longer be exfiltrated. The sanction against the telecom operators Free and Free Mobile, in January 2026, shows this clearly. The intrusion covered 24 million subscriber contracts, including some subscribers' IBANs. Among the failings identified, Free Mobile had kept “millions of items of its subscribers' data, without justification, for an excessive period”.
That finding is separate from the intrusion itself. Yet it says the essential thing: data that has become useless keeps creating risk for as long as it exists. Hence retention periods set by type of data, inactive accounts deleted or anonymized, and exports that expire.
Try a simple exercise: if someone exported your application's data tomorrow, what would they find? How many years of history? Which documents? And what could you have deleted long ago?
Author · M.Baudoux - Software development & AI consulting
Sources
- Cybermalveillance.gouv.fr, "What to do in the event of a cyberattack? (Guide for business leaders)" (updated March 17, 2026)
- CNIL, "Personal data breaches: the rules to follow"
- Cybermalveillance.gouv.fr, "Data breaches in 2025: an acceleration with very diverse consequences" (June 1, 2026, updated August 18, 2026)
- CNIL, "Cybercrime: risks and consequences for personal data" (November 26, 2025, Harris Interactive survey of 2,082 people)
- ANSSI, "Threat summary on generative AI in cyberattacks" (February 4, 2026)
- DGSI, Economic interference flash no. 117, "Risks associated with the use of artificial intelligence in the professional world" (December 2025)
- CNIL, sanction against Hôpital privé de la Loire (September 3, 2026)
- CNIL, sanctions against Free Mobile and Free (January 13, 2026)
- CNIL, "Data retention periods" (updated April 2, 2026)
