Cybersecurity and AI: turn investment into operational readiness
Market attention to cybersecurity is widening the discussion about how businesses handle digital risk. Leaders can turn that attention into accountable decisions, practical tests and recovery criteria.
Leggi in portogheseBy Pentesty
From market attention to a business decision
In its November 19 report, Brazil Journal connects advances in AI, digital dependence and investor interest in cybersecurity. That discussion can help bring the subject to senior leadership. Pentesty’s analysis builds on it to consider how businesses can turn attention into operational decisions.
For a business, the useful question is which capability needs to improve: preventing unauthorized access, identifying an incident, containing its reach or restoring an activity. An outage requires investigation before its cause can be classified as an attack. Predictions about AI also need to remain separate from confirmed incidents and results observed within the organization’s own environment.
Start with the activity that must continue
The NIST CSF 2.0 organizes security outcomes to support risk assessment, prioritization and communication. Organizations of different sizes can choose how to achieve those outcomes. In the practical approach proposed here, leaders begin with a critical activity, such as billing, customer service or releasing a logistics operation.
Record who owns that activity, which systems and providers support it, and how much downtime the business can manage. Include shared dependencies such as identity, connectivity and support. Two separate contracts may still be exposed to the same failure. Our analysis of connection route diversity explains why continuity planning needs to examine the actual underlying dependency.
Give AI a task and an access boundary
The NIST AI RMF is a voluntary resource for considering risks when developing, using and evaluating AI systems. Its page states that version 1.0 is being revised. For each business application, we propose documenting its task, accessible data, credentials and actions that require human approval.
An assistant that searches documents and one that changes permissions need assessments suited to their roles. Before integration, test whether an out-of-scope action is refused, a permitted action is recorded and access can be revoked. Use an authorized environment and appropriate data. Our analysis of network boundaries for agents develops this connection between capability and authorization.
Fund response and recovery work
The NIST incident response recommendations, published in April 2025, incorporate preparation, detection, response and recovery into risk management. They provide a basis for discussing the work that accompanies a technology purchase: available owners, rehearsed procedures and clear conditions for resuming an activity.
Arrange a controlled exercise involving operations and security. Check who receives the alert, who decides to interrupt access and how the team communicates if its usual channel is unavailable. During recovery, request evidence that the critical function works again and its access permissions have been checked. Record elapsed times, failed dependencies, test limitations and outstanding corrective work.
Bring verifiable outcomes into the budget
Leadership can track each initiative through five details: the risk addressed, the activity affected, its owner, its deadline and the evidence required for acceptance. Examples include verified removal of unnecessary access, a fix validated in production or a restoration observed during an exercise. Teams should explain the scope of each outcome and the risks that remain open.
This discipline also helps compare proposals. Alongside price, examine deployment effort, staff availability and the provider’s participation in testing. A first cycle can focus on one activity and produce a record of decisions and outstanding work. The program can then expand using what the organization has learned, with explicit criteria for its next investment.
Frequently asked questions
How should the first investment be prioritized?
Choose a critical activity, identify a relevant risk to its operation and decide who can accept the outcome. Document the expected evidence before comparing solutions.
Does an outage prove that an attack occurred?
Its cause needs investigation. Record the confirmed impact, available evidence and unresolved hypotheses, updating communications as the investigation progresses.
What should be checked before connecting an AI agent?
Check its task, data, credentials and permitted actions. Test access boundaries and revocation within an authorized scope, recording decisions and observed limitations.
How should a recovery exercise be assessed?
Observe restoration of the critical activity, measured times, access checks and required dependencies. Record what the exercise did not cover and corrective work requiring follow-up.
Sources and references
- Brazil Journal: reporting by Matheus Prado, November 19, 2026
- NIST: Cybersecurity Framework 2.0, published February 2024
- NIST: AI Risk Management Framework, page consulted November 19, 2026
- NIST: incident response recommendations, SP 800-61 Rev. 3, April 2025
Original Pentesty analysis, published November 19, 2026. The proposed scenarios are editorial guidance, with scope and validation to be defined for each environment.