人工智能时代的网络韧性适应性

Age of AIEnsure sovereignty and trust at AI speed C-Suite briefCybersecurity has reached an inflection point, driven by a fundamental change in its operating conditions. The erosion of traditional security perimeters has been underway for years – driven by mobility, remote work,globalization, cloud, and digital ecosystems. What has changed isspeed and scale. Threats now operate atmachine speed,outpacing defense models shaped largely around human-led responses.AI is reshaping both offense and defense. Regulations demand continuous, provable execution, not periodicassurance. And sovereignty has moved from policy debate to operational requirement. Cyber risk is now the number one global business risk, placing it firmly at the top of the board agenda*. Yet evenas cybersecurity budgets increase, scrutiny is intensifying because risks, costs, and expectations are outpacinginvestment. The question for leadership is no longer whether cybersecurity activity is increasing, but whether theorganization can regain control quickly when disruption occurs, and prove it. The cybersecurity architectures most enterprises rely on today are built on sound principles, but were designedfor execution at human speed, across fragmented environments, and proven largely through documentation.That model is no longer sufficient as IT and OT converge, AI and data pipelines become operational dependencies,ecosystems expand, and disruption propagates across dependencies faster than humans can coordinate. Adding tools and controls hasn’t solved the execution gap; what closes it is a change in operating discipline.Closing the gap requires shifting cybersecurity from prevention-first control accumulation to a resilience operatingdiscipline that assumes disruption and focuses on limiting impact and recovering fast. Atos Group calls thisAdaptive Cyber Resilience. It reframes cybersecurity as anadaptive operating disciplinethat is focused on maintaining control, recoveringfast, and continuously proving resilience under pressure rather than proposing yet another framework, checklist,or product stack. This model brings together four inseparable shifts: •AI embedded at the core of security operations•Sovereignty treated as a design principle•Security measured by business outcomes rather than tool counts•Human accountability retained even when execution operates at machine speed Adaptive Cyber Resilience is where speed meets discipline. Through a continuous cycle ofprepare, respond,and adapt,it turns resilience from a goal into a governed operating discipline.Preparebuilds readiness beforedisruption.Respondcontains impact and enables engineered, demonstrable recovery at speed.Adaptensuressecurity evolves faster than the environment it protects. A governance control layer connects board-level intent This paper outlines what has changed, the questions leaders must confront, and a concrete path to regainingcontrol through disciplined execution. It is written for executives accountable for outcomes, not experimentation. Atos Group brings more than 25 years of experience operating in mission-critical environments, sovereigninfrastructure, and industries where systems must operate under pressure, fail safely, and withstand scrutiny.Through Atos Cyber Services and Eviden Cyber Products, this experience operates as one integrated system atscale, under pressure, and over time. The shift: Why the oldmodel breaks Every board recognizes cybersecurity as a strategic risk. Fewer can answer a simple question:If something goeswrong tomorrow, how quickly do we regain control, and can we prove it? That gap between recognition and readiness is not a failure of effort. Controls still matter. Compliance still matters.But static controls and periodic assurance no longer match today’s operating reality. Business now runs through constantly changing dependencies across cloud, IT, OT, partners, APIs, data flows,and AI systems. This extends exposure beyond the boundaries most security models were designed to manage,while creating an opportunity to rethink controls for greater efficiency. The result is compounding risk that is ofteninvisible, rarely owned end-to-end, and increasingly difficult to contain. The traditional perimeter has faded. Today’s effective perimeters are identities, APIs, and increasingly themodels and agents that operate across ecosystems. Control is distributed, dependencies run deep, and manyorganizations no longer fully map the environments they must secure. Here’s why: attacks and connected OT expands the controlgap, security must be judged by its ability to detect,contain, and recover at speed. The issue is no longertechnology alone. It is an operating model that is no 1.AI changes the equation - permanentlyAI does not simply accelerate cybersecurity; it changes the control model. As attacks movefaster and weaknesses surface sooner, human-led operations struggle to absorb the pace.At the same time, agentic systems introduce 4.Th