By Rekha Shenoy, CEO, BackBox
Mythos’ AI-powered vulnerability discovery capabilities have demonstrated that flaws in your software can be found faster than ever and chained into working exploits. The number of common vulnerabilities and exposures (CVEs) has increased steadily over the years. However, in the AI era, CVEs are on track to exceed 60,000 by the end of 2026, a 25% increase from 2025. Hackers can now exploit vulnerabilities within days or hours of disclosure and routinely before a patch is released.
When a patch or workaround is available, it’s important to act fast, but even large organizations struggle to keep up. According to the Verizon 2026 Data Breach Investigations Report, only 26% of critical vulnerabilities, defined by CISA’s known exploited vulnerabilities (KEV) catalog, were fully remediated by organizations in 2025, down from 38% the previous year, and the median time to full resolution increased by nearly two weeks, from 32 to 43 days.
These trends are particularly concerning for network administrators. A CISO at a large insurance company told me they have a vulnerability remediation tool for servers and workstations that store sensitive data. Unfortunately, when it comes to network infrastructure, they can’t adequately scale their team, so they accept the risk to network devices. They wait for a hardware refresh or apply software updates when possible, even though the cost of downtime has just reached $15,000 per minute.
With Mythos and other AI-capable models in the picture, speed is now the new risk. Network administrators must shift from manual patching to a broader set of automated and AI-enabled, proactive best practices to strengthen the overall security posture. Instead of buying into hype, let’s be pragmatic about how we address the challenge so we can begin to operate at machine speed. The following recommendations can help:
Understand your organization’s true risk from vulnerabilities. Attackers organize themselves like businesses and emphasize return on investment, so only a small fraction of CVEs are actively exploited in the wild. In 2025, this amounted to 238 high-risk vulnerabilities that CISA added to its KEV catalog. Even with AI, what works in a test environment is very different from what works in the real world. For a vulnerability on a network device to be worthwhile to exploit, it needs to be internet-facing, prevalent across organizations, and exploitable in a repeatable manner. Context-enriched vulnerability prioritization enables you to focus on mitigating those that truly expose your organization to risk.
Build cybersecurity resilience by reinforcing the basics. Kobe Bryant used to say, “I never get bored with the basics.” The best basketball player in the world attributed his success to routine drills he repeated consistently during warmups. Similarly, when NetOps teams prioritize first principles recommended by CIS, they can build a superior security control program and secure their business with minimal risk. These principles include asset inventory and control, security configuration, account management, access control, continuous vulnerability management, and disaster recovery. Many CIS controls should be checked and reinforced daily, but often aren’t due to the same manual labor problem that hinders timely vulnerability remediation. Solving this problem is very doable and is now more urgent.
Lay the groundwork with automation. Organizations are pushing their teams to do more with less by using AI rather than hiring more staff. However, if you aren’t using automation, you’re putting the cart before the horse. AI requires context about your network devices to be useful. How can you manage vulnerabilities or check configurations for compliance if you don’t have a current inventory of devices? This includes device manufacturer, type, model number, software and firmware versions, and configuration. Automation lets you aggregate, store, and update your device inventory in real time, enabling better control over assets. Another basic use case that supports CIS controls is automating backup and restore, so you always have the latest version in the event of a crisis. These two use cases are essential building blocks for AI-enabled security configuration, continuous vulnerability management, and disaster recovery.
Incorporate AI in a trustworthy way. Use validated data and best practices when incorporating AI to drive efficiencies, reduce errors, and increase reliability. For example, AI can quickly discover, aggregate, and normalize vulnerability data from trusted sources, including CISA, NCD, NIST, and device manufacturer websites, then map that information to your device inventory and prioritize remediation based on risk scores. You can use AI similarly when running compliance checks to identify and flag configuration drift against CIS benchmarks and internal policies.
Automate complex workflows at scale. Leverage agentic NetOps to chain automations together and help streamline complex workflows such as performing upgrades or modifying configurations across a wide fleet of network devices. AI acts as an informed assistant to run device checks, generate tasks, gather vendor recommendations and internal best practices, test automated workflows, and report issues needing human review. Teams can move from reactive fixes to proactive risk management.
Our response to Mythos should not create more chaos and uncertainty by introducing more agents that bring bad with the good. We need to respond with the basics: focus on first principles, address the human capital problem with automation and AI-driven intelligence that uses trusted data sources and validated workflows, and then implement agentic NetOps with human-in-the-loop oversight to move even faster at scale.
