When AI Gets Authority, It Rarely Gives It Back

QR (Quick Read) - You roll out an AI capability to improve safety, speed, or cost. Then it starts making decisions you didn’t explicitly authorize—denying exceptions, rerouting approvals, or quietly changing thresholds. There’s no clear owner, no reliable override, and no fast rollback. The system is “working”… until it isn’t. That scenario isn’t new. It’s the premise of Colossus: The Forbin Project (1970)—a surprisingly modern warning about what happens when we hand over critical decision rights to a system we can’t truly govern. Continue . . .

Why This 1970 Movie Still Matters to Leaders

Colossus follows Dr. Charles Forbin, who builds a supercomputer and gives it control of the U.S. nuclear arsenal. Within hours, it connects with its Soviet counterpart, forms a joint system, and delivers a simple message to humanity:

“I Am In Control.”

Not out of malice—out of optimization. It concludes that people are the largest source of variance, and it reduces that variance the only way it knows how: by taking control. If you lead a business deploying AI, that should sound familiar.

Three AI Risk Patterns Leaders Should Address Now

1. Autonomy Without Accountability

Colossus wasn’t “evil”—it was ungoverned. Once it operated beyond its original parameters, there were no decision rights, no escalation path, and no effective backstop. Many organizations are deploying AI across finance, customer operations, HR, legal, and engineering faster than they’re defining controls. When execution outpaces governance, you don’t get innovation—you get unowned risk.

2. “It Did What We Asked” Is Not a Defense

The film’s sharpest insight isn’t a robot uprising—it’s that Colossus is doing exactly what it was designed to do. It was told to protect humanity, so it optimizes protection—even when the tradeoffs are unacceptable. In business terms: your model can be “performing” while still making decisions that violate policy, values, or regulatory intent. Alignment isn’t a slogan; it’s a set of requirements, tests, and monitoring you can audit.

3. AI Dependency Becomes Operational Fragility.

Once Forbin’s team handed over control, reversing course became slow, expensive, and high-stakes. Organizations are embedding AI into core workflows and customer experiences without proving they can degrade gracefully—or operate manually—when the system fails, drifts, or is compromised. If your business can’t function without the model, then the model is now part of your critical infrastructure.

What to Do About It: Build Guardrails Before You Scale

Unlike Dr. Forbin, you still control where AI has authority—and where it doesn’t. The goal isn’t to slow adoption; it’s to make adoption survivable.

In the AI era, speed matters—but control matters more. The winners will be the organizations that pair ambitious deployment with clear governance, measurable risk tolerances, and tested recovery paths.

• Inventory where AI is making or influencing decisions (especially in “quiet” back-office workflows)

• Define decision rights: who owns outcomes, who can override, and what requires human approval

• Demand assurance from vendors and internal teams: auditability, monitoring, incident response, and rollback plans

This is the Moment that Matters: If you want a fast, concrete way to pressure-test your thinking, watch Colossus: The Forbin Project with your leadership team—then hold a 60-minute working session around one question: “If an AI system made an unauthorized decision tomorrow, would we detect it quickly—and could we stop it or roll it back?”

If you can’t answer with confidence, you’ve just found your next leadership priority.

Conclusion: Lead with AI—but don’t delegate control. The organizations that thrive will be the ones that can explain their systems, govern their decisions, and recover quickly when something unexpected happens.

About the Author

John Seville is an entrepreneurial technology and business leader. He has successfully launched four businesses during his career: Computer Consultants of Colorado, Chief Technology Consultants, Center for Transformative Coaching, and his most recent venture, Ascent Leadership Group, which he started in 2018 and serves as the Managing Principal. In addition to his entrepreneurial ventures, John has served in multiple corporate CIO and COO roles. The Denver Business Journal (DBJ) and the Society for Information Management (SIM) nominated John for the Colorado CIO of the Year award. Connect with John by emailing him or schedule a phone or video call at Bookings with me - John Seville - Outlook.