What this week’s cyberattacks tell us about control effectiveness

Compromised website software and autonomous attacks raise a familiar question: How do we know our cybersecurity controls work?

A person visits a familiar website. The organization has a recognizable name, an established business, and presumably a cybersecurity program. The visitor has little reason to suspect that code arriving through that website could attempt to steal payment information or compromise a device. Yet reports published this week describe precisely that risk.

Researchers identified malicious JavaScript associated with websites belonging to 178 organizations through the Optimizely web experimentation platform. The affected scripts targeted visitors to sites associated with organizations including CNN, Zoom, and Salesforce. Optimizely confirmed that an unauthorized party exploited a vulnerability in its platform to inject malicious code into customer projects. The company said it fixed the vulnerability and removed the code. The extent of successful compromises remained unclear in the initial reporting.

Separately, Google’s Threat Intelligence Group reported an AI framework that could identify software vulnerabilities, develop exploits, and execute attacks without human intervention. Google also described broader adoption of agentic tools that accelerate stages of an intrusion while continuing to exploit familiar weaknesses, including compromised credentials and unpatched systems.

The reports describe different attack methods. Together, they illustrate the pressure facing cybersecurity programs: organizations must understand an expanding network of dependencies while defending against attackers who can operate with increasing speed and scale.

For cybersecurity professionals, risk managers, and internal auditors, the practical question remains familiar.

What evidence gives us confidence that our controls can address the risks we face?

A trusted service can become a source of exposure

Organizations rely on third-party services to support ordinary business activities. A website might use external tools for analytics, payment processing, customer support, advertising, or testing different page designs.

Each integration serves a purpose. Each also creates a dependency that someone needs to understand and manage.

The Optimizely incident illustrates how a compromise involving a shared service can reach visitors across many organizations. Attackers can gain considerable reach by targeting software that multiple businesses trust and embed in their websites.

Consider what that means for a typical vendor review. An organization might confirm that a provider has a security policy, obtain an assurance report, review contractual language, and approve the relationship. Those activities provide useful information. Their value depends on whether the organization connects them to the way it actually uses the service.

For a service that delivers code to a website, the assessment needs to consider what that code can do, which pages load it, what information those pages handle, and how the organization would respond to suspicious activity.

A procurement record may identify the vendor. A technical inventory should help explain the exposure.

A completed questionnaire may describe the provider’s security practices. Monitoring and testing should help establish whether the organization’s own safeguards operate as intended.

The practical lesson is to follow the dependency into the business process. Understand what the organization has allowed the service to do and what could happen if someone abuses that capability.

Faster attacks put pressure on control timing

The Google report adds another dimension: speed.

Google described attackers using AI and agentic frameworks to automate work across stages of an intrusion. Its analysts emphasized that familiar weaknesses remain relevant even as attackers gain new ways to exploit them more quickly.

That distinction matters. Discussions about AI can pull attention toward extraordinary scenarios while unresolved vulnerabilities, excessive access, and exposed systems continue to create opportunities.

Organizations need to understand how faster exploitation affects the time available for their controls to work.

A vulnerability management process, for example, might meet its documented remediation deadlines. Management could receive a dashboard showing that most patches fall within the required window. An auditor could verify those figures and conclude that the process operates consistently.

The next question concerns the design of that window.

Does the process escalate vulnerabilities affecting exposed, critical systems? Can the team respond when evidence of active exploitation changes the urgency? Who can authorize temporary restrictions when a permanent fix needs more time?

A control can operate exactly as documented and still provide insufficient protection if its design assumes more response time than the organization actually has.

The same concern applies to monitoring. An alert has limited protective value if nobody reviews it promptly, understands its significance, or has authority to act.

Control timing belongs in the effectiveness conversation.

Evidence should show what the control accomplished

Internal auditors already know how to ask the questions these developments require.

When we examine an access review, we want to understand what the reviewer checked, whether the information was complete, and what happened when access needed to change. A signature helps establish that someone performed an activity. We still need evidence that the activity addressed the risk.

Cybersecurity control testing requires the same discipline.

Suppose an organization monitors changes to scripts on important website pages. A configuration screenshot may show that someone enabled the monitoring capability. Stronger evidence would help us understand whether it covers the relevant pages, identifies meaningful changes, routes alerts to the appropriate people, and supports a timely response.

Suppose management reports that it remediated a critical vulnerability. A closed ticket records the decision. Verification should establish that the fix reached the affected systems and that the exposure no longer exists.

The objective is to connect evidence to the result the control should produce.

Documentation supports that work. The conclusion needs to reach beyond the existence of documentation.

Cybersecurity, risk management, and internal audit need a shared view

Incidents involving shared services and faster exploitation also expose the limitations of working in functional silos.

Cybersecurity teams understand technical dependencies, defensive capabilities, and operational constraints. Risk managers help connect those issues to business priorities and risk tolerance. Internal auditors independently evaluate whether controls address the risks and operate as intended.

Each function contributes something different. Their work becomes more useful when they share a clear understanding of the business exposure.

Take a customer-facing website that depends on several external services. Cybersecurity may identify a suspicious script change. Business leaders need to understand whether disabling the service would interrupt transactions. Risk management can help evaluate that tradeoff. Internal audit can later assess whether the organization had clear responsibilities, reliable information, and an effective response process.

Management owns the decisions and controls. Internal audit provides independent assurance over them.

Keeping those responsibilities clear allows the functions to coordinate without blurring accountability.

Use the headlines to ask better questions

A major incident creates an opportunity to examine our own assumptions.

For a shared-service compromise, we can ask where we depend on similar integrations, who maintains the inventory, and how quickly the organization could identify and restrict an affected service.

For faster vulnerability exploitation, we can examine how management changes priorities when new threat information arrives. We can also look at whether remediation reporting reflects the exposure of critical systems or hides it inside broad completion percentages.

For both scenarios, we can ask management to walk through the evidence.

Show us how an alert reaches someone who can act. Show us how a remediation decision becomes a verified fix. Show us how the response team identifies the business processes at risk.

Those conversations can reveal gaps that a policy review alone would miss. They can also demonstrate where controls work well and give leadership a sound basis for confidence.

No control environment can guarantee that an organization will avoid every attack. Effective controls should help it reduce exposure, detect problems, limit damage, and recover. Testing should help leaders understand how much confidence they can place in those capabilities.

Why I wrote Cybersecurity Controls

The gap between having a control and knowing whether it works is a central theme of my upcoming book, Cybersecurity Controls: Design, Implement and Audit to Protect Your Organization.

I wrote it for professionals who need to connect cybersecurity risks to practical action. The book covers administrative, physical, and technical controls, along with approaches for evaluating their design and testing their operating effectiveness.

Cybersecurity teams need controls they can operate consistently. Risk managers need a clear understanding of the exposure those controls address. Internal auditors need evidence that supports meaningful conclusions.

Recent headlines make that work feel urgent. The need will continue long after attention moves to the next incident.

Organizations will keep adopting new technology and depending on external services. Attackers will keep looking for ways to exploit weaknesses faster. Our responsibility is to keep examining whether the control environment can meet those changing demands.

Choose one critical business process and trace its cybersecurity controls from the stated objective to the evidence of performance. Where does your confidence hold up, and where does it depend on an assumption?

Cybersecurity Controls releases December 3 in Europe and December 29 in the US.

Explore the book and preorder from Kogan Page. Use promo code KOGANPAGE25 for 25% off.

Next
Next

Fraud Red Flags vs. Fraud Indicators: Why the Difference Matters for Auditors