The picture that emerges is of a faster and higher-volume threat, without evidence, for now, that it is systematically more effective. Authorities and companies still face the same imperative: patch flaws quickly, even as the tools capable of searching for them spread.
Since at least August 2026, hackers have been using Artex, a free vulnerability-scanning tool based on autonomous artificial-intelligence agents, against organizations worldwide, according to the FBI.
The FBI report, published on November 13, cites targets in financial services, public administrations, technology, defense, semiconductors, energy and health care.
This spread illustrates a shift already observed: advances in AI also serve cybercriminals, who use them as a force multiplier to carry out hacking on a larger scale.
Google says it detected, for the first time in the last quarter of 2026, malicious actors capable of fully automating the discovery and exploitation of vulnerabilities, without human intervention. French targets are among the organizations concerned, according to these findings.
France, like other countries, faces numerous and dangerous cyberattacks, from thefts of personal data to hacks targeting everyday equipment, including supermarket refrigerators.
AI does not necessarily transform how attacks themselves work. It mainly changes their pace: when one path fails, a system can test another within a second, whereas a conventional hacker might take several hours.
This speed also makes it possible to try to exploit vulnerabilities that have only just been made public. The volume of attempts can therefore rise before organizations have had time to patch their systems.
AI-augmented attacks are not perfect, however. In a report, the company Citalid says that, among the cyberattacks studied, the use of AI still shows neither a general increase in the success rate, nor an increase in the severity of payloads, nor a lasting structural advantage for the attacker.
Citalid adds that no end-to-end autonomous operation has been documented. The full automation described by Google therefore remains, at this stage, a warning signal rather than an established norm.
Recent examples of AI-assisted hacking have nevertheless made the news. Several cyberattacks presented as carried out autonomously with the help of AI from OpenAI and Anthropic made headlines this summer.