Le Journal

The predictions around AI in 2026 from Neil Roseman, CEO, Invicti. AI will continue to do impressive, strange and dangerous things in 2026: Agentic AI will likely be the dominant trend, and we'll hopefully see even more complicated applications of the technology including the rise of agent-to-agent systems. With the growth of AI agents, AI security will be affected both for the good and bad. We'll probably see major breaches related to AI, as the very presence of the technology allows attackers to do things faster, more easily and in new ways. That goes for AI technologies - such as agents - which can be used to malicious ends, as it does for attempting to compromise the AI technologies that targets use. Attackers will continue to try and poison the AI supply chain and look for vulnerabilities in existing models, which could give them huge capabilities to use against their targets. AI security, however, will likely make huge leaps too. There are a whole range of security functions such as analytics or software code analysis that AI will likely take over, with more purpose-built domain specific models emerging to cover specific functionalities. In turn, that may help fill the yawning skills gap that constantly dogs security departments. After 2024 and 2025's waves of redundancies around the tech world, security teams are being asked to do more with less. AI security technologies - especially agentic - technologies may ease much of that pressure and allow them to multiply their effect with limited resources and manpower. That said, some part of the AI bubble will likely burst in the coming year. There is simply too much money flowing into too many companies and we'll likely see an overdue correction in 2026. However, unlike the dotcom bubble of the early 2000s, there is real value here and despite a burst in some parts of the market, AI will continue to grow in the coming years. Data centers will continue to be a thorn in the side of the AI industry, as AIs growth outstrips the ability for infrastructure to scale with it. Data center capacity has been under immense pressure in the last few years as a direct result of the growth of AI. This is partially fuelled by AI's enormous need for power and cooling, which the industry still struggles to meet. That need, however, could be met by technological solutions or regulatory changes which might prompt innovation in the area.

An attacker can traverse directories of Django, via archive.extract(), in order to write a file outside the service root path. View online : https://vigilance.fr/vulnerability/...