A few years ago, the hardest question in cybersecurity was how to keep pace with an increasingly sophisticated threat landscape. Today, the question is simpler: What happens when the threat can move faster than the organization defending against it?
We now have a concrete answer. Earlier this year, Anthropic disclosed an AI model, Claude Mythos Preview, that could autonomously read a codebase, find previously unknown vulnerabilities, and write working exploits for them, at a scale and speed no human team could match.
The company judged the capability dangerous enough that it declined to release the model to the public. Within weeks, comparable capabilities surfaced from other labs. The offensive side of security now operates at machine speed, and that capability is proliferating.
The defensive side, for the most part, doesn’t.
For decades, organizations measured remediation—finding a flaw, fixing it, and proving it was fixed—in weeks. A vulnerability was discovered, a ticket opened, team assigned, release scheduled, evidence gathered, and the issue eventually closed.
That cadence made sense when defenders had time. They no longer do. Across the industry, the window between a vulnerability becoming public and its first exploitation has collapsed from months to days and, for the most serious flaws, to hours.
Many exploited vulnerabilities are now weaponized within a day of disclosure, and a growing number before any patch exists at all. When exploitation is measured in hours, a remediation process measured in weeks is not a security strategy. It’s a compliance exercise. And compliance, on its own, has never stopped an attack.
AI exposes an outdated operating model
Most executives I speak with already understand that AI is changing their business. But AI also exposes the limits of how their organizations operate. For years, the response to security pressure has been to add more tools, dashboards, processes, and people.
Despite record spending, most enterprises still run workflows built for a slower era, and the remediation sequence feels normal because it has been for a long time. It’s also increasingly too slow to matter.
This is a failure of design. The operating model was never built to compete with a machine-speed adversary, and no amount of effort within that model can close the gap.
Lost innovation as hidden cost
While most conversations about security focus on risk, the more revealing conversation is about productivity. Every board wants faster product delivery and a better return on its technology investment.
Yet in large enterprises, thousands of engineering hours each year go to keeping existing software current and patching libraries, updating frameworks, managing dependencies, remediating findings, and maintaining evidence for auditors. While required, very little of this work differentiates the business.
That creates an innovation debt. Every hour an engineer spends manually keeping technology up to date is an hour not spent on a customer, product, or new capability. The organizations must free their best people from the maintenance treadmill.
Compliance is becoming continuous
Regulators have been signaling this shift for years, and AI is accelerating it. Compliance was historically retrospective. Gather documentation, prepare for the audit, and demonstrate that a process is in place.
Technology now changes continuously with every deployment, dependency, and day. Examiners increasingly expect organizations to demonstrate ongoing control rather than proving you were compliant on the day of the audit.
The future of compliance is proving, continuously, that it works. That requires continuous visibility, evidence, and remediation, not a quarterly fire drill.
You can’t hire your way to machine speed
When organizations realize they’re falling behind, the instinct is to add headcount. The volume of change is greater than any team can absorb manually when modern enterprises run thousands of applications, often built on code they didn’t write and don’t control.
Past a certain scale, adding people adds coordination overhead with more handoffs, meetings, and approvals, faster than it adds throughput. The economics don’t work, the speed doesn’t work, and the risk doesn’t work. At enterprise scale, automation is now a business requirement.
Machine-speed remediation, with human judgment
Organizations must stop treating remediation as a sequence of human tasks and start treating it as a self-running system. Machines identify the issue, generate the fix, produce the evidence, and validate compliance. People stay firmly in control of governance, risk decisions, and exceptions, but they are no longer the ones performing repetitive work at scale.
This is the same pattern that reshaped finance, manufacturing, and customer service. In each case, the industries redesigned how work was done rather than working harder. Security is now on the same path and must operate at machine speed while keeping human oversight where it matters most.
None of this is theoretical. My teams operate exactly this model for one of the largest U.S. banks, across more than 20,000 repositories (the separate codebases the large bank runs) and roughly 2,000 engineering teams. Most remediations now close the same day they are raised, and 99.5% land by their compliance due date.
Changes that once took a quarter happen in days. In one case, a migration across 5,500 repositories completed in two. The system has returned thousands of engineering hours to product teams and pays back roughly $4 for each it costs to run. That is what defending at machine speed looks like in practice.
The real competitive advantage
Security has been treated as a defensive function. It’s necessary but separate from growth. That framing is becoming obsolete. An organization that can continuously keep its technology current will adopt AI, modernize, ship, respond to threats, and pass audits faster, and will have more engineering capacity left over for the work that actually moves the business.
A security capability becomes a business capability. That, in the end, is the broader lesson AI keeps teaching leaders. Rethink the systems and operating models surrounding systems to achieve stronger security and speed.
When the exploit window is measured in hours, every enterprise faces the same choice: Operate at machine speed, or accept machine-speed risk.
Chris Brown is the managing director at Sngular US.