A new front line in the artificial intelligence war
On 14 April 2026, OpenAI launched GPT-5.4-Cyber, a model specialising in cybersecurity, just days after Anthropic announced Claude Mythos. This timing is no coincidence: it marks an acceleration in the competition between AI giants on what has become critical ground, the security of IT systems. In a world where cyberattacks now cost more than 10 trillion dollars a year globally, cybersecurity is becoming a major strategic issue for businesses1. AI is no longer just a productivity tool; it is becoming a pillar of defence.
Models able to understand attacks, not just detect them
GPT-5.4-Cyber does not merely analyse logs or apply conventional detection rules. It draws on advanced contextual understanding to identify abnormal behaviour, correlate events and simulate attack scenarios. This approach makes it possible to move from a detection mindset to one of anticipation. In practice, the model can analyse network traffic, detect weak signals and suggest corrective actions in near real time. This development is essential at a time when the volume of security data is exploding, making traditional approaches inadequate.
A technology race between OpenAI and Anthropic
The launch of GPT-5.4-Cyber is part of a direct rivalry with Claude Mythos. Both models aim to automate part of cybersecurity to improve the speed and accuracy of responses. But beyond the technology, this showdown reflects a broader battle for control of critical infrastructure. AI players are seeking to become indispensable partners for businesses in managing their systems, data and security. Cybersecurity is thus becoming a strategic lever in the competition between AI platforms.
Significant operational gains for businesses
For businesses, and SMEs in particular, these tools represent a major opportunity. Cybersecurity is often complex and costly to implement, especially for organisations with limited resources. Models such as GPT-5.4-Cyber make it possible to automate certain tasks, such as monitoring, analysis or incident prioritisation. According to IBM, using AI can cut data breach costs by up to 40% and significantly speed up response times2. This allows businesses to strengthen their security without proportionally expanding their IT teams.
Growing dependence on intelligent systems
However, this automation brings new risks. Businesses are becoming increasingly dependent on AI models whose inner workings they do not always understand. Misinterpretations, biases or false positives can have serious consequences. Moreover, the concentration of these tools in the hands of a few players raises questions of sovereignty and data control. Cybersecurity no longer depends solely on internal infrastructure, but also on the AI platforms used.
Cyberattacks are also being supercharged by artificial intelligence
One of the major issues with this development is that attackers are using AI too. Cybercriminals exploit models to automate the creation of malware, improve phishing attacks or bypass detection systems. This is creating a genuine technological arms race, in which every defensive advance triggers an offensive adaptation. GPT-5.4-Cyber and Claude Mythos are part of this dynamic: they strengthen defences, but also help raise the overall level of sophistication of attacks.
Cybersecurity and AI: a new generation of IT skills
This transformation highlights a key skill: cybersecurity, now inseparable from artificial intelligence. IT professionals must understand how models work, know how to integrate them into an overall architecture and be able to interpret their results. It is no longer just about protecting systems, but about managing intelligent tools capable of evolving continuously. This hybridisation is redefining IT jobs and the digital skills expected by businesses.
A revolution under way, but still far from settled
GPT-5.4-Cyber illustrates a profound transformation of cybersecurity, but this revolution is still taking shape. While AI can improve threat detection and response, it does not replace human expertise. Businesses must learn to use these tools with discernment, combining automation and oversight. For students and future IT professionals, understanding this duality is essential: tomorrow’s cybersecurity will rest on a balance between human intelligence and artificial intelligence, in a constantly changing environment.
360° IT News — decoding technology to understand the jobs of tomorrow.
References
1. Cybersecurity Ventures. (2024). Cybercrime damages report.
https://cybersecurityventures.com
2. IBM. (2024). Cost of a Data Breach Report.
https://www.ibm.com/security/data-breach
$32 billion: Google launches a cyber army of AI agents
Cybersecurity enters a new dimension Google has made a major move by announcing an investment of $32 billion in developing a “cyber army” of artificial intelligence agents. Behind this term lies a technological reality: systems capable of automatically detecting,...
Data centre boom: Legrand goes on the offensive in China and the UK
Data centres are becoming the heart of the global digital economy Legrand is stepping up its investments in China and the United Kingdom to strengthen its presence in the data centre sector, which has become critical infrastructure for the digital economy. This...
Claude Mythos: beyond the buzz, what are the real stakes for IT security?
An AI that promises to transform cybersecurity With Claude Mythos, Anthropic is showcasing a new generation of artificial intelligence able to analyse, detect and anticipate cybersecurity threats. At a time when cyberattacks are soaring, with more than 2,200 attacks a...
Microsoft reinvents development with AI agents in VS Code
Software development enters the age of agents Microsoft is taking a new step by integrating artificial intelligence agents into Visual Studio Code, with a beta version available via the Insiders channel. This development comes at a time when VS Code is used by more...
In the cloud era, enterprise networks are changing standards
The network is no longer just support: it is becoming a driver of performance For a long time, the enterprise network was regarded as a simple technical infrastructure, often invisible as long as it worked. But with the explosion of the cloud, SaaS applications and...
Windows 11: Microsoft wants to speed up user testing with Insider
Software testing becomes a lever for innovation Microsoft is overhauling its Windows Insider programme, which now brings together more than 10 million testers worldwide, to simplify access to experimental versions of Windows 11 and speed up user feedback. In an...
51,600 satellites: Jeff Bezos steps up the battle for the global internet
News in brief With Project Kuiper, Jeff Bezos is committing Amazon to a major transformation of the world’s internet infrastructure by deploying tens of thousands of satellites in low Earth orbit. The goal is not limited to connecting isolated areas, but to building a...
PepsiCo automates its logistics: a turning point for smart warehouses
News in brief PepsiCo France is investing in an automated, multimodal warehouse that combines robotics, digital systems and intelligent management of logistics flows. The aim is no longer simply to store and distribute products, but to turn the warehouse into a...
Windows 11: Microsoft finally relaxes forced updates
News in brief Microsoft is preparing to overhaul the way updates are managed in Windows 11, giving users greater control over their installation, notably by making it easier to postpone them or avoid certain enforced interruptions. This change responds to a...
WeChat goes autonomous AI: Tencent steps up with OpenClaw
News in brief Tencent has just integrated OpenClaw into WeChat, turning the app into far more than a simple communication tool. This development introduces agents capable of carrying out tasks directly from the interface, without going through several applications or...
