Reading · The Streetlight

In 2023, UnitedHealth was sued over the use of an algorithm to deny care for Medicare Advantage members. The lawsuit alleges patients were told their post-acute care was over because an algorithm said so, even though the plan’s denials were reportedly overturned in more than 90% of appeals.

Welcome to the world of AI dominance. Not the demos, the hype, or the “superintelligence” bedtime stories. A world where a machine can cut you off in seconds, and then it is on you to fight your way back through appeals, paperwork, and years of expensive legal pressure, after the damage has already landed.

This is how the modern system works when it has permission. It runs on your time. On delay, confusion, and exhaustion. It turns stolen human effort into profit, and then tells you it is “efficient.”

When a system can hurt people cheaply and nobody pays for it, it keeps doing it.

AI did not invent that system. We did. AI just makes it cheaper, faster, and easier to hide.

AI risk is not identical everywhere. Tools are powerful, but the surrounding system decides how that power is directed, managed, and whether it becomes help or harm.

AI becomes most dangerous when three conditions stack together.

A small number of people can impose a change across everything quickly. Less friction. Fewer veto points. Fewer humans allowed to say no. Speed can build good things fast. Speed can also spread a bad decision across millions of lives before anyone can stop it.

When oversight is weak, corruption grows. Secrecy and opacity become normal. Procurement happens behind closed doors. “Emergency” becomes the excuse. “Trust us” becomes the policy. Untested systems end up installed in the highest-stakes domains because there is no serious cost for getting it wrong.

Some systems treat people as expendable in practice. Names in a spreadsheet. Data points. Targets. Harm can be ignored, delayed, blamed, buried, or normalized as the cost of doing business.

Stack those three and AI becomes an accelerant. A black box where everything is reduced to inputs and workflows, justified as efficiency, and hidden inside financial statements.

When harm is cheap and nobody pays for it, it becomes policy.

That is why AI is not going to fix a broken system by default. It will follow the incentive path. It will do what we allow it to do.

You do not need futurists to imagine the future. The pattern is already around you.

A class action claims Cigna relied on an automated algorithm to reject health insurance claims without doctors properly reviewing patients’ files. The reporting and allegations describe PxDx being used to deny claims in bulk as “not medically necessary,” with denials happening at machine speed.

Even if you argue about details, the structure is the warning.

Companies will use automated systems because they are cheap and fast. Those systems will amplify the existing harm those companies were already causing.

The machine does not need to hate you. It only needs a goal, a budget, and the lack of consequences.

We are sold a utopia story. But these systems are amplifiers of the world we already built.

They are not autonomous operators connected to some higher truth. We set the objective. We choose the tradeoffs. We decide what counts as “success.” We press the go button.

And when it goes wrong, blame dissolves.

The corporation points at “the system.” The system points at “the model.” The model points at “the data.”

You are left arguing with fog.

There is a popular argument going around that AI will hit society the way COVID hit society: suddenly, everywhere, and bigger than anyone expects.

The warning about impact is fair, but the metaphor is wrong.

A virus does not ask permission. AI is software. Software does not spread on its own. It needs contracts, procurement, budgets, leadership decisions, and legal cover.

It spreads where it is installed, where it is connected, and where it is allowed.

COVID revealed institutional character. In places where truth became tribal and media incentives turned safety into a culture war, correction got harder and harm got easier.

We were powerless against a virus. But we are not powerless against software.

AI works inside information systems. We design and manage these systems. We choose what they optimize. We choose whether there is transparency and accountability.

If your society cannot tell the truth about what is happening, it cannot fix what is happening. It will spin. It will deny. It will protect image over people.

If those systems are redesigned to serve people, AI can be a boon. If we keep the profit-over-people mindset and growth at any cost, AI will make the same harms cheaper and more scalable.

The “AI alarm” people are sounding is just one more siren in a world already screaming about climate catastrophe, inequality, scams, corruption, racism, fascism, war, and genocide.

AI did not create the rot. It exposes it. It accelerates it.

The internet keeps telling one story as if the world is one country. It is not.

Around the world, culture and community can act like guardrails. Leaders fear backlash. Employers live inside communities. Courts work. Labor can organize. Journalists can investigate. Rules create friction.

Slow implementation can function as safety because it gives people time to see what is being installed and fight back.

The US is a special risk case for structural reasons.

The UN General Assembly just approved a 40-member scientific panel on AI impacts and risks, with the United States strongly objecting.

The US also sits high on measures of government AI readiness, meaning it has strong capacity to roll AI into public services and systems.

It also has massive private-sector deployment power, and a political-media environment that can turn safety into a team sport, which breaks the correction loop when things go wrong.

Put it together and you get the dangerous mix unfolding in plain view.

Any analysis that tells you to just “adapt” while executives and politicians wire AI into the systems that decide whether you get care, housing, education, or a living is corrosive.

It is surrender dressed up as realism.

Yes, prepare yourself. Learn the tools. Build skills. Do not lie to yourself about the labor market. But do not accept the scam that personal excellence is enough.

If a system is built to treat people as disposable, your resume will not save you when it is profitable to discard you.

You cannot solve a structural problem with personal responsibility. Self-improvement is not enough. We need public standards, enforced.

If AI touches health, welfare, justice, policing, education, housing, immigration, or anything that decides who eats and who suffers, these are non-negotiables:

Disclosure: The public has the right to know where AI is used in high-stakes decisions.

Testing before scale: No mass deployment without rigorous evaluation and published results.

Appeals that work: A real human must be able to reverse the decision quickly, with clear timelines.

Incident reporting that cannot be buried: The OECD has already pushed a common reporting framework for global benchmark for stakeholders across jurisdictions and sectors.

Named responsibility: One accountable owner for outcomes, not a fog of vendors and “the algorithm.”

If an institution refuses these, they are telling you the truth about who they are and how they operate. They want the power without the consequences.

We have been told the people at the top are too smart, too rich, too powerful to challenge. They want you quiet and tired. They want you to believe nothing you do matters.

That story is a tool of control.

You are not helpless. You are being trained to feel helpless.

So raise your voice. Organize. Boycott.

Pressure your city, your school board, your hospital system, your insurer, and your lawmakers.

Support the journalists who expose this. Support worker groups that can slow reckless deployments. Build communities that make harm very expensive, and disposability a crime.

Because this is still our world.

The only question that matters is whether we live in a society where harm is cheap and nobody pays for it, or a society where every deployment comes with a name, a decision is logged in a record, and every action comes with an enforceable consequence.

This is our fight. And it is winnable.

AI Is Not a Virus

Comments