AI Is Compressing the Cybersecurity Timeline. Security Teams Are Racing to Keep Up.

Key points

• Anthropic, OpenAI, and DeepSeek have each released major AI model advances in recent months, putting capabilities within reach of security teams and criminals alike.
• Security work that once took specialist analysts weeks, such as finding software flaws or mapping attack paths, can now run in hours.
• Chief information security officers, the executives responsible for an organisation's digital safety, have shifted their main question from "Can AI help?" to "How fast can we deploy it?"
• The bottleneck in security operations is no longer finding threats. It's acting on what you already know, quickly enough to matter.
• AI can expand what a small analyst team can investigate, but only if the system is tuned to the specific environment it's protecting.

For years the big question in cybersecurity boardrooms was whether artificial intelligence could actually do useful security work. Most organisations treated AI as a future project, interesting but not urgent.

That framing has quietly collapsed.

Security executives speaking with CSO Online now describe a different pressure. They're not debating whether AI works. The question is how much time they have left before falling behind.

The trigger is a run of high-profile model releases. Anthropic's Project Glasswing and its Mythos model, together with OpenAI's Daybreak system and advances from the Chinese AI lab DeepSeek, have each demonstrated that AI can now reason through problems that previously required years of specialist training. Finding weaknesses in software, mapping how an attacker could move through a network, crafting convincing fake messages to trick employees: these tasks are becoming faster and cheaper for anyone with access to capable AI. Our 20 July story on Ivanti showed what that looks like in practice, when the company used AI to uncover a perfect-score flaw in its own software.

How does this affect ordinary people and the organisations that serve them?

When criminals work faster, the window between a break-in and real damage shrinks. A flaw in a hospital's booking system or a retailer's payment platform that might once have taken attackers weeks to find and exploit could now surface within days. The cost of slow patching or delayed responses rises sharply as a result.

Inside security teams the change is already visible. Security operations centres, the teams of analysts who monitor for intrusions around the clock, were built on a simple idea: alerts arrive, humans investigate, and you hire more analysts to handle more alerts.

AI changes that equation. An investigation that requires checking dozens of systems across endpoint devices, email and network logs can now run in minutes rather than hours. The analyst's role shifts toward directing the investigation, evaluating its output, and hunting for threats the automated system might miss.

Moving fast without tuning your tools is its own risk. A system that treats every organisation's network as identical will produce results nobody can verify. The teams seeing real gains are those that configure their AI tools to their own specific environment before trusting the output.

Should you worry about phishing?

For employees at any organisation, the practical concern is harder-to-spot phishing. Criminals send fake messages to trick staff into handing over passwords or clicking malicious links, and AI can now personalise those messages at scale. Healthy scepticism about unexpected requests for credentials or urgent action has become a baseline habit, not an occasional reminder.

Security teams aren't losing this race yet. But the margin for slow decisions is getting smaller, and the pace of model releases suggests it will keep shrinking.