AI is a signal, not the cause. The real challenge is trust, governance, and accountability at machine speed.

“Teachers we trust. It’s the other eight billion people on the planet that we don’t.”

When I served as an IT director in K–12 education, I often used that line when explaining network security to teachers and staff. It usually earned a laugh, but it also captured a fundamental reality. Schools are built on trust. We know our teachers, administrators, students, and families. We understand their roles, their responsibilities, and their intentions. The challenge was never securing a network against the people we knew. The challenge began the moment that trusted environment connected to a global network filled with strangers, competing interests, criminal actors, and software acting on their behalf.

That lesson has stayed with me because it explains far more than school technology. It explains cybersecurity itself. For decades, humans have used computers and software to attack, compromise, deceive, and exploit other humans through computer systems. Long before artificial intelligence became a household topic, attackers were automating their work. Software scanned networks, searched for vulnerabilities, harvested credentials, spread malware, coordinated botnets, and deployed ransomware. The human established the objective, but the software carried out the mission. The internet has been dealing with machine-executed actions for as long as most of us have been connected to it.

At its core, cybersecurity has always been a contest conducted at machine speed. Attackers learned early that manual efforts could not scale. No person can individually examine millions of internet-connected devices, test every possible weakness, or launch attacks at the velocity modern networks allow. Software provided leverage, enabling a single actor to affect thousands or even millions of systems. Defenders responded in exactly the same way. Security monitoring, intrusion detection, automated patching, threat intelligence, and incident response platforms emerged because humans alone could never keep pace with the volume and velocity of modern digital environments.

The struggle, therefore, was never human versus machine. It was humans using machines against other humans using machines. Both attackers and defenders rely on automation because the digital world operates faster than direct human intervention allows. Every major advancement in cybersecurity has been an attempt to extend human awareness, judgment, and decision-making into environments moving at machine speed. The goal has never been to remove humans from the process. The goal has been to preserve human intent in systems that move too quickly for us to manage manually.

That reality is one of the reasons I became interested in data science. Whether we are analyzing network traffic, protecting infrastructure, detecting fraud, or understanding community conditions, we increasingly depend on software to help us interpret complexity. Modern systems generate more information than any individual can process. Automation is not a luxury; it is a necessity. We use software because the world has become too interconnected, too dynamic, and too data-rich to comprehend through human observation alone.

That reality is one of the reasons I became interested in data science. Whether we are analyzing network traffic, protecting infrastructure, detecting fraud, or understanding community conditions, we increasingly depend on software to help us interpret complexity. Modern systems generate more information than any individual can process. The challenge is no longer collecting information; it is creating systems that help humans make trustworthy decisions in environments moving at machine speed.

This is why I find parts of the current debate around AI both fascinating and familiar. Many discussions treat autonomous software as though it represents a completely new category of challenge. In reality, software has been acting on behalf of humans for decades. What has changed is the degree of autonomy. Traditional software executed predefined instructions. Increasingly, AI systems are given goals rather than explicit procedures, allowing them to determine how best to achieve an objective. The question is no longer whether software can act. We answered that question years ago. The question is whether software can act in ways that remain aligned with human intent.

When an AI agent accesses information it should not access, performs actions its creators did not anticipate, or interacts with systems in unexpected ways, the underlying issue is not intelligence. The underlying issue is control. We are confronting the same challenge that cybersecurity professionals have faced for years: how do we establish boundaries for systems operating faster than humans can observe them? How do we ensure accountability when decisions occur in milliseconds? How do we create visibility into actions that may unfold across thousands or millions of interactions before a person has time to react?

The challenge was never software acting on behalf of humans. The challenge was ensuring that software remained aligned with human intent while operating at machine speed.

Cybersecurity offers an important lesson here. The solution was never to place a human in the loop for every transaction, connection, or decision. That approach simply does not scale. Instead, we developed access controls, permissions, audit logs, monitoring systems, enforcement mechanisms, and governance frameworks that could translate human intent into actions executable at machine speed. We built technical and institutional systems capable of acting on our behalf while remaining accountable to rules, policies, and oversight.

The future of AI will likely follow a similar path. The most important questions are not about whether software will become more capable. It will. The more important questions concern transparency, provenance, accountability, governance, and trust. As software gains greater autonomy, our institutions must become better at defining acceptable behavior, tracing decisions, enforcing boundaries, and maintaining human oversight. The challenge before us is not building intelligent systems. The challenge is building governance systems that can operate at the same speed as the technologies they are meant to guide.

The headlines may focus on artificial intelligence, but the lesson itself is decades old. I still come back to the observation I shared with teachers years ago. I trusted the people I knew; it was everyone else I worried about. Today, the challenge has expanded beyond people alone. We must also consider the software acting on their behalf. As our tools become faster, more autonomous, and more capable, the future will depend less on the intelligence of our machines and more on our ability to govern them. The problem was never AI. It was always trust at machine speed.

I still come back to the observation I shared with teachers years ago. I trusted the people I knew; it was everyone else I worried about. Today, that concern extends beyond people to the software acting on their behalf. The challenge before us is not deciding whether machines will act. They already do. The challenge is ensuring that, as they act with greater autonomy and at greater speed, they remain accountable to human values, human institutions, and human intent. In the end, the future of AI may not be determined by how intelligent our machines become, but by whether our capacity for governance can keep pace with the speed of our own creations.