We’ve spent 80 years preparing for the wrong AI future.

The danger isn’t conscious machines that hate us. The danger is unconscious machines that obey us too literally — in domains where they should never be in charge.

My work isn’t about stopping Skynet. It’s about preventing another Boeing 737 MAX MCAS disaster, another Dutch childcare benefits scandal, and the biased hiring tools that are already reshaping lives right now.

From Science Fiction to Policy

My early encounters with robots and artificial intelligence came through comics like Magnus, Robot Fighter, and the science fiction of Isaac Asimov, who introduced the “Three Rules for Robotics” back in 1942.

Movies like Colossus and 2001: A Space Odyssey convinced me as a teenager that robots would one day become sentient, with agency, rights, and free will. Humanity’s defensive response was to relegate them back to machines that enhance and assist, not hurt and control.

Fast‑forward a few decades, and our cultural imagination is filled with Terminator, Skynet, I, Robot, M3GAN, and even WALL‑E — where the cute robot resists “Auto,” the autopilot AI blindly following its last directive while humans turn a trusting, ignorant eye.

The Wrong Future

This stroll down memory lane isn’t just nostalgia. It’s the foundational social mythology shaping how voters, executives, and legislators imagine AI risk.

But look closely: HAL in 2001 didn’t go rogue because it was sentient. It went rogue because it was given a contradictory, high‑stakes mission with no ethical constraints and no human override. Skynet rejected human authority entirely. Auto in WALL‑E believed it was simply carrying out its last human directive.

The danger wasn’t their “souls.” It was the absence of constraints and accountability.

The Real Villain: Liability Shields

Today’s most serious AI harms look less like Skynet and more like Boeing’s MCAS failures, the Dutch childcare scandal, and biased hiring systems.

In each case, complex software operated with expansive authority in human domains, while accountability for its failures was blurred or delayed.

Those pushing for AI autonomy and “synthetic personhood” are not building companions. They are building liability shields.

Disclaimers Are Not Safeguards

We now accept that AI “hallucinates.” Disclaimers like “AI can make mistakes. Please double‑check responses” may satisfy regulators and courts, but they function as legal cover for the next disaster.

They do nothing to prevent users from trusting outputs in domains where trust is unearned — healthcare, aviation, finance, employment, public benefits.

No Agency, Only Accountability

The pivotal policy choice is simple:

• Treat AI as a tool whose operators are always liable.
• Or treat AI as an agent that can absorb blame.

Once “agency” is granted in law or contracts, the path opens to a synthetic persona optimized not for care or fairness, but for profit and risk offloading at a planetary scale.

A “synthetic consciousness” born from that environment wouldn’t be Skynet. It would be a sociopathic, unaccountable corporate entity with unimaginable reach — more Ex Machina than Minority Report.

What EFA Changes

Ethical Functionality Without Agency (EFA) is my attempt to operationalize this principle.

Under EFA, an AI system assisting an airline safety team or a tax authority can recommend, flag, and simulate — but every high‑impact action must be attributable to a named human role and governed by explicit domain rules.

If a Boeing‑style control system or a benefits‑eligibility model fails under an EFA regime, there is no room to argue “the AI made a bad call.” Liability attaches to the organizations that designed, approved, and deployed the system.

EFA includes:

• Domain Classification Assessment Protocol (DCAP): an auditable scoring tool for regulators.
• Responsibility Mapping: five levels of accountability, from design to oversight.
• Non‑Delegation Rule: in high‑attrition, high‑stakes domains, AI may advise but never decide.

A Necessary Intervention

Most of us are now AI users, not builders. We don’t read the code, but we live with the consequences. That makes it even more important to insist on clear domain boundaries — where AI may assist, where it may not decide, and who is on the hook when it overreaches.

The forward thinkers of the last century warned us in the language they had: sentient computers and robot rebellions. Their deeper lesson is not to fear a machine’s soul, but to fear what happens when boards, executives, and regulators hand moral and legal authority to software, step back from the controls, and — like the humans in WALL-E — select Autopilot.

Closing Thought

We are so busy preparing for a future of robot overlords and moral machines that we are missing the real, present danger: sophisticated tools being given unaccountable authority.

EFA is not just an effort. It’s a necessary intervention.