Cybersecurity teams face an overwhelming number of new vulnerabilities. The volume of threats now exceeds their capacity to respond, a challenge made worse by shrinking budgets and a shortage of skilled professionals worldwide.
Dan Elliott, Field CISO for APJ at Recorded Future, stated that AI-driven attacks have caused a five-fold increase in reported Common Vulnerabilities and Exposures (CVEs). Following recent releases like Mythos and Glasswing, Mozilla rolled out patches en masse and Palo Alto Networks reported a 5x jump in discovered vulnerabilities.
More threats, fewer hands to fix them
The National Vulnerability Database recorded about 50,000 CVEs in 2025. Yet fewer than 1% were actively exploited, according to data from Recorded Future. This discrepancy between possible and actual threats has led security leaders to adjust their strategies.
Elliott explained that budget constraints and staffing shortages create difficult choices. Teams require tools that separate meaningful threats from background noise, focusing only on vulnerabilities already being used by attackers, not just those with high severity scores.
The change involves more than speed. AI models, whether used for harm or unintentionally, now run on affordable hardware, making attacks easier to launch. Defensive AI tools are also being deployed for automated threat detection, helping short-staffed teams manage the workload.
From detection to business risk
A financial services firm recently redesigned its vulnerability management by combining Recorded Future’s intelligence with its IT service tools. The automation reduced manual review time by over 20 hours weekly, allowing the team to concentrate on higher-risk issues.
Elliott believes threat intelligence often remains confined to security operations. When shared with executives, it can guide broader decisions—supporting budget requests, influencing procurement, and aligning cybersecurity with business goals. Too often, these insights stay locked in dashboards that leadership never reviews.
Related: NSW to use facial recognition in gaming
AI can now simulate attacks to test whether patched systems would withstand real-world threats. This capability may connect immediate detection with long-term risk management. The difficulty lies in integrating these tools while preserving human oversight.
Elliott clarified that AI does not replace people. It expands their ability to process data and focus on threats that truly endanger assets. The objective is to transform raw information into practical guidance for executives and boards.
Achieving this change requires more than software. It involves rethinking priorities—moving from reacting to every alert to identifying which risks matter most to the organization. Partners, Elliott added, should help teams value accuracy over features, ensuring tools match real-world dangers.
The pressure is increasing. Threat actors may not be more sophisticated, but AI allows them to operate faster. Security leaders must determine not whether they can fix every vulnerability, but whether they can identify and address the ones that will be exploited before damage occurs.
Conditions are evolving rapidly. The focus must remain on immediate priorities and tools that predict the next move.
Elliott noted that cosmic mysteries may seem distant from cybersecurity, but both fields rely on advanced technology to filter vast amounts of data and highlight what truly matters.
