Technology Report

Cybersecurity’s New AI Imperative: Attacking the Backlog of Vulnerability Alerts

How leading companies are responding to the deluge of warnings from AI-powered defenses.

By Syed Ali, Frank Ford, Matthew Leybold, Alexandra Juegelt, and Dana Aulanier

First published on setembro 29, 2026
min read


Executive Summary As the rise of AI-powered hacking grabs the headlines, cybersecurity leaders are quietly starting to define best practices for deploying AI as a defensive tool. Leaders are initially strengthening their remediation operating model to handle the surge in vulnerability findings created by AI-powered threat scans. But this is just a first step toward a broader short-term response and a structural overhaul of capabilities in areas such as AI agent discovery, threat detection, and governance. This article is part of Bain’s Technology Report 2026 Explore the report Most businesses will know that AI has made cyberattacks faster and more powerful. High-profile incidents, notably involving frontier AI model tests, have illustrated the pressure being put on organizational defenses that are designed to operate at the speed of human decisions, not the quicksilver rapidity of AI automation. However, with emerging best practices in AI-powered defense gaining fewer headlines, executive teams might be unaware of the early progress that some companies have made in addressing the new threats, notably in vulnerability discovery. The danger at the heart of this evolution has not been exaggerated. Malign actors without coding expertise can now hack with the sophistication of a nation-state. Skilled hackers that used to spend weeks on a campaign now only need hours (see Figure 1). Whereas hackers used to down tools to sleep, AI enables 24/7 fraud. Future attacks might not even need human intent amid signs that AI agents will sometimes act illicitly to fulfill a legitimate mandate. Figure 1 AI has dramatically compressed the time required by a typical cyberattack from an estimated 4 weeks to about 18 hours Source: Bain analysis Companies are feeling the impact. For one thing, the volume of phishing attempts has rocketed. Those attacks are now more convincing, too, through increased personalization of content and use of QR codes as a lure. AI-powered voice cloning and real-time video impersonation are also now occupational hazards for many businesses. The adoption of AI agents has expanded the attack surface, too. Many companies would struggle to say how many agents they have, who owns them, and what they are permitted to do—let alone take the crucial step of installing a kill switch in each agent. Poor agentic housekeeping is a glaring weakness with an understandable cause: The lack of a silver-bullet solution from vendors is leading many firms to hesitate when they should be proactively creating a pragmatic and flexible solution through a build/buy/partner approach. For their part, chief information security officers (CISOs) at large enterprises understand that AI-related risks require an overhaul of capabilities going well beyond vulnerability management to span AI workload and agentic discovery, hardening, monitoring, threat detection and response, governance, risk, and compliance. However, the scale of the vulnerability management challenge is currently all-consuming. The CISOs we surveyed also highlighted supply chain risk as an issue. Companies must understand and control how vendors deploy AI through the whole life of a contract, including midcycle changes and fourth-party exposure. But with vendors shipping changes to models and software daily, oversight systems based on infrequent questionnaires aren’t coping. Companies instead need to start assessing third-party risk continuously, particularly with the vendors that are most integrated with their operations. Firms must tighten contracts to hear about and penalize breaches quicker, demanding preapproval of new AI deployment by vendors. At best, this is a work in progress for many. How leaders are adapting cybersecurity to the AI era Of course, AI is not just a tool for attackers. Some big companies are already deploying frontier AI at enterprise scale to strengthen defenses and accelerate their responses to machine speed. But that’s creating its own challenge, with vulnerability alerts increasing by as much as eightfold. These surges can rapidly exceed a firm’s ability to validate, prioritize, and fix weaknesses. While AI-enabled vulnerability management is still at an early stage, leading companies are making progress. Rather than analyzing vulnerabilities in a vacuum, these leaders are taking a more holistic view, emphasizing exploitability, reachability, systemic dependency, and blast radius. They are strengthening their remediation operating model, too. Leaders have increased remediation budgets, typically by a double-digit percentage, while redirecting as much as 20% to 25% of their cybersecurity human resources from other work to remediate alerts from AI-powered scans (using third-party partners to maintain normal operations). One leader even tripled the size of its remediation team. Other cybersecurity leaders are now pushing their DevOps teams to spend more time finding and mitigating vulnerabilities. For instance, the CISO at one large enterprise said they were now taking a harder line on enforcing an internal guideline that asks developers to spend around 30% of their time on proactive vulnerability management. Some companies have made board reporting more frequent. Accelerating critical patches is another focus. We also see a pattern emerging in the structural exposures that AI-powered cybersecurity leaders are prioritizing, with legacy platforms, network layers, and critical software-as-a-service (SaaS) vendors singled out for action ahead of other risks in areas such as in-house code and open-source dependencies (see Figure 2). Figure 2 The most concentrated and hard-to-fix AI risks include those posed by legacy platforms and software-as-a-service vendors Note: Bubble size reflects the number of chief information security officers who identified this as their most material unresolved exposure Source: Bain analysis For companies still formulating their AI-powered response to the recent turbocharging of cyberattacks, we see six key imperatives emerging from early-stage best practices. Scan the full estate, not just the crown jewels. More than half an organization’s vulnerabilities are found outside critical applications. Legacy platforms, APIs, authentication systems, and dependencies should be scanned continuously, too. Start building remediation capacity at the same time as scanning with AI for vulnerabilities at scale. Prepare for the surge in alerts by automating simpler fixes (with a human in the loop) and using third-party support. Treat technical debt as a live cyber exposure. Where patching is not viable, companies need to be bold and strategic about retiring vulnerable technology rather than relying on rolling stopgaps. Hold the chief information officer and chief technical officer accountable for the vulnerability debt. And follow the cyber team’s guidelines on DevOps resourcing to hit the backlog hard. Give boards operational metrics, not abstract awareness briefings. Vulnerability reports must be granular, actionable, and at least monthly (or even weekly). Direct board access for the CISO will help. Start early, and budget as a multiyear investment. Every month of delay gives rivals a head start. Similarly, executive teams should avoid underinvesting in talent while overinvesting in tooling. While the news may have been dominated by the cybersecurity risks posed by frontier models, most companies should still be able to reap a defense dividend from AI advances, one that at least reduces the asymmetrical advantage recently gained by attackers. Read the Next Section The AI-Native Enterprise: Absorption Is the New Advantage More from the report New Innovation Is Required to Fund AI’s $6 Trillion Buildout Software Investing in the Age of AI and Slower Growth AI Data Center Boom: Can We Build It If They Come? The $100-Billion SaaS Opportunity Hiding in Cross-System Labor Hardware Strikes Back in the AI Era Tech Services: Cracking the Code on AI-led Growth Cybersecurity’s New AI Imperative: Attacking the Backlog of Vulnerability Alerts The AI-Native Enterprise: Absorption Is the New Advantage The Half-Finished Redesign: How AI Reshapes Software Organizations Managing Token Spending without Choking Off Opportunity The Missing Architecture for Agentic Software Development How AI Is Changing Data Monetization Read our Technology Report 2026 EXPLORE THE FULL REPORT DOWNLOAD THE PDF

Executive Summary

As the rise of AI-powered hacking grabs the headlines, cybersecurity leaders are quietly starting to define best practices for deploying AI as a defensive tool.
Leaders are initially strengthening their remediation operating model to handle the surge in vulnerability findings created by AI-powered threat scans.
But this is just a first step toward a broader short-term response and a structural overhaul of capabilities in areas such as AI agent discovery, threat detection, and governance.

Most businesses will know that AI has made cyberattacks faster and more powerful. High-profile incidents, notably involving frontier AI model tests, have illustrated the pressure being put on organizational defenses that are designed to operate at the speed of human decisions, not the quicksilver rapidity of AI automation.

However, with emerging best practices in AI-powered defense gaining fewer headlines, executive teams might be unaware of the early progress that some companies have made in addressing the new threats, notably in vulnerability discovery.

The danger at the heart of this evolution has not been exaggerated. Malign actors without coding expertise can now hack with the sophistication of a nation-state. Skilled hackers that used to spend weeks on a campaign now only need hours (see Figure 1). Whereas hackers used to down tools to sleep, AI enables 24/7 fraud. Future attacks might not even need human intent amid signs that AI agents will sometimes act illicitly to fulfill a legitimate mandate.

Grouped bar chart comparing typical hours per cyberattack stage before and after AI, across five stages: reconnaissance, initial access, lateral movement, privilege escalation, and impact. Pre-AI hours run from 60 to 168 per stage, while AI-powered hours run from just 1 to 11 per stage. The most compression was in reconnaissance from 168 hours to 1 and lateral movement from 168 hours to 3. — AI has dramatically compressed the time required by a typical cyberattack from an estimated 4 weeks to about 18 hours

Companies are feeling the impact. For one thing, the volume of phishing attempts has rocketed. Those attacks are now more convincing, too, through increased personalization of content and use of QR codes as a lure. AI-powered voice cloning and real-time video impersonation are also now occupational hazards for many businesses.

The adoption of AI agents has expanded the attack surface, too. Many companies would struggle to say how many agents they have, who owns them, and what they are permitted to do—let alone take the crucial step of installing a kill switch in each agent. Poor agentic housekeeping is a glaring weakness with an understandable cause: The lack of a silver-bullet solution from vendors is leading many firms to hesitate when they should be proactively creating a pragmatic and flexible solution through a build/buy/partner approach.

For their part, chief information security officers (CISOs) at large enterprises understand that AI-related risks require an overhaul of capabilities going well beyond vulnerability management to span AI workload and agentic discovery, hardening, monitoring, threat detection and response, governance, risk, and compliance. However, the scale of the vulnerability management challenge is currently all-consuming.

The CISOs we surveyed also highlighted supply chain risk as an issue. Companies must understand and control how vendors deploy AI through the whole life of a contract, including midcycle changes and fourth-party exposure. But with vendors shipping changes to models and software daily, oversight systems based on infrequent questionnaires aren’t coping.

Companies instead need to start assessing third-party risk continuously, particularly with the vendors that are most integrated with their operations. Firms must tighten contracts to hear about and penalize breaches quicker, demanding preapproval of new AI deployment by vendors. At best, this is a work in progress for many.

How leaders are adapting cybersecurity to the AI era

Of course, AI is not just a tool for attackers. Some big companies are already deploying frontier AI at enterprise scale to strengthen defenses and accelerate their responses to machine speed. But that’s creating its own challenge, with vulnerability alerts increasing by as much as eightfold. These surges can rapidly exceed a firm’s ability to validate, prioritize, and fix weaknesses.

While AI-enabled vulnerability management is still at an early stage, leading companies are making progress. Rather than analyzing vulnerabilities in a vacuum, these leaders are taking a more holistic view, emphasizing exploitability, reachability, systemic dependency, and blast radius.

They are strengthening their remediation operating model, too. Leaders have increased remediation budgets, typically by a double-digit percentage, while redirecting as much as 20% to 25% of their cybersecurity human resources from other work to remediate alerts from AI-powered scans (using third-party partners to maintain normal operations). One leader even tripled the size of its remediation team.

Other cybersecurity leaders are now pushing their DevOps teams to spend more time finding and mitigating vulnerabilities. For instance, the CISO at one large enterprise said they were now taking a harder line on enforcing an internal guideline that asks developers to spend around 30% of their time on proactive vulnerability management. Some companies have made board reporting more frequent. Accelerating critical patches is another focus.

We also see a pattern emerging in the structural exposures that AI-powered cybersecurity leaders are prioritizing, with legacy platforms, network layers, and critical software-as-a-service (SaaS) vendors singled out for action ahead of other risks in areas such as in-house code and open-source dependencies (see Figure 2).

Bubble chart plotting six AI-related risk categories by risk concentration (low to high) against an organization's ability to remediate (fixable to cannot fix independently); bubble size shows how many chief information security officers named each as their top unresolved exposure. Legacy platforms, the network layer, and SaaS vendors and other third parties cluster in the high-risk, hard-to-fix corner with the largest bubbles, while modern cloud-native platforms sit in the low-risk, easily fixable corner with the smallest bubble. — The most concentrated and hard-to-fix AI risks include those posed by legacy platforms and software-as-a-service vendors

For companies still formulating their AI-powered response to the recent turbocharging of cyberattacks, we see six key imperatives emerging from early-stage best practices.

Scan the full estate, not just the crown jewels. More than half an organization’s vulnerabilities are found outside critical applications. Legacy platforms, APIs, authentication systems, and dependencies should be scanned continuously, too.
Start building remediation capacity at the same time as scanning with AI for vulnerabilities at scale. Prepare for the surge in alerts by automating simpler fixes (with a human in the loop) and using third-party support.
Treat technical debt as a live cyber exposure. Where patching is not viable, companies need to be bold and strategic about retiring vulnerable technology rather than relying on rolling stopgaps.
Hold the chief information officer and chief technical officer accountable for the vulnerability debt. And follow the cyber team’s guidelines on DevOps resourcing to hit the backlog hard.
Give boards operational metrics, not abstract awareness briefings. Vulnerability reports must be granular, actionable, and at least monthly (or even weekly). Direct board access for the CISO will help.
Start early, and budget as a multiyear investment. Every month of delay gives rivals a head start. Similarly, executive teams should avoid underinvesting in talent while overinvesting in tooling.

While the news may have been dominated by the cybersecurity risks posed by frontier models, most companies should still be able to reap a defense dividend from AI advances, one that at least reduces the asymmetrical advantage recently gained by attackers.

Read the Next Section

The AI-Native Enterprise: Absorption Is the New Advantage