Does AI Replace the Security Leader?
Why Command Means "You Decide."
One of the loudest promises in security today is that AI will run the program on its own: triage, decide and act. For a leader accountable to a board, that promise is as much a threat as a solution. If the machine decides, who is accountable when it is wrong? The discipline AQUILA C4I is built on answered that question long ago.
What C4I Was Designed to Do
The National Research Council's 1999 review of U.S. defense C4I programs defined C4I systems as systems designed to support a commander's exercise of command and control. Support is the operative word. C4I exists to give the person in command the clearest possible picture, so that the decision, and the accountability, stay with that person. The same report noted that its principles apply to large, complex private enterprises as much as to defense.
Where AI Already Pays
AI is not optional in security, and it is already delivering. IBM's 2025 Cost of a Data Breach study found that organizations using security AI and automation extensively shortened their breach lifecycle by 80 days and saved nearly $1.9 million on average. The global average cost of a breach fell 9 percent, largely because breaches were identified and contained faster. Leaders are responding: PwC's 2026 Global Digital Trust Insights found that AI enablement of key cyber capabilities is the top priority for cyber budgets.
Where AI Creates New Risk
The same IBM study shows the other side. Thirteen percent of organizations reported breaches of AI models or applications, 97 percent of those lacked proper AI access controls, and 63 percent of organizations had no AI governance policies. Shadow AI, used without approval or oversight, was involved in 20 percent of breaches. AI adopted without governance has become a new source of the risk it was meant to reduce.
Advising and Acting Are Different Designs
| Question | AI That Acts | AI That Advises |
|---|---|---|
| What it asks | What can I do without the person? | What does the person need to see to decide? |
| Where accountability sits | Unclear, and often nowhere | With a named person, on the record |
| What happens when it is wrong | The error is already in production | A person catches it before it is acted on |
| What a board or auditor hears | "The system decided" | "Our leader decided, with the evidence, and here is why" |
| Where it fits best | Narrow, reversible, well-bounded tasks | Decisions with business, legal or regulatory weight |
Matching Autonomy to Consequence
The two designs are not opposites. The right question for any task is how much autonomy its consequences allow.
| Task | Consequence If Wrong | Appropriate Role for AI |
|---|---|---|
| Enriching and grouping alerts | Low, and easily corrected | Act automatically |
| Isolating a single endpoint | Moderate, and reversible | Act within set limits, with notice to a person |
| Ranking the day's decisions | Moderate, because attention is finite | Recommend, with the reasoning shown |
| Accepting a risk or funding a control | High, and lasting | Advise only; a person decides |
| Disclosing an incident | High, legal and regulatory | Draft only; a person decides and signs |
Why Accountability Cannot Be Delegated
Regulation assumes a human answer. Since 2023, U.S. public companies have had to disclose material cybersecurity incidents within four business days of determining they are material, and to describe each year how their board oversees cyber risk and what expertise management brings to it. A disclosure, a risk acceptance or a board report needs a person who made the decision and can explain it. AI can make that person faster and better informed. It cannot stand in for them.
Five Questions Before Adopting Security AI
- Which decisions will it make on its own, and which will it only recommend?
- Who reviews its recommendations, and is that review recorded?
- Can it show where each figure came from, and say when it cannot see something?
- What access does it have, and who governs that access?
- If it is wrong, how quickly will a person know?
Conclusion
AI does not replace the security leader. It changes where the leader's time goes: less of it spent assembling the picture, more spent deciding. In AQUILA C4I, AQUILA AI drafts and ranks; a named person reviews and approves every advisory; and the organization owns the decision. The machine advises. You command.
Sources
- National Research Council, Realizing the Potential of C4I: Fundamental Challenges, 1999
- National Research Council, 1999, Executive Summary
- All Covered, on AI savings in IBM's 2025 report
- Help Net Security, on AI-related breaches in IBM's 2025 report
- Bluefin, on AI governance and shadow AI in IBM's 2025 report
- PwC, 2026 Global Digital Trust Insights
- U.S. Securities and Exchange Commission, press release 2023-139
