AI Could Cure Cancer or Cement Autocracy. Which Way It Goes Is Up to Us

Key points

• Katherine Rundell, author and Oxford academic, warns that AI is already shifting wealth and power toward a small number of individuals.
• She argues AI could be used to spread disinformation, enable mass surveillance, and run weapons that kill without human pilots or soldiers.
• She also lists genuine benefits: faster cancer research, better disaster warnings, wider access to legal help.
• Her central claim, published in The Guardian, is that who controls AI matters more than what AI can do.

Two futures live inside the same technology. That is the uncomfortable argument Katherine Rundell, author and academic, puts to readers this week, and it is hard to dismiss.

On one side: a tool that could, in her words, speed the cure of cancer, help disabled people live more independently, and give ordinary citizens the kind of legal knowledge that currently costs hundreds of pounds an hour. Early-warning systems for floods and wildfires could get sharper. Scientists could find patterns in Earth's ecosystems that no human team could spot alone.

On the other side sits something darker.

What is Rundell actually warning about?

Her sharpest concern is concentration of power. AI, she writes, is already enabling a "breathtaking shift" of wealth from the many to a tiny group of men whose attitude toward human life she finds alarming. She does not name them, but the point lands.

Beyond wealth, she flags three specific dangers. First, disinformation at a scale no previous technology could match: AI can generate fake text, images and video without armies of human workers behind it. Second, mass surveillance, the ability of governments or corporations to watch and record citizens so completely that dissent becomes dangerous. Third, lethal autonomous weapons, military systems that select and kill targets without a human making the final call.

Taken together, she argues, these capabilities describe something close to a ready-made toolkit for authoritarian rule.

Does this mean AI is bad?

Rundell says no, but she sets a hard condition. The benefits only materialise if the people controlling the technology have something like good values. If AI ends up concentrated in the hands of "crooks and despots", the medical breakthroughs become irrelevant.

That is not a technical argument. It is a political one.

Her focus is education: classrooms are where the next generation of voters, engineers and policymakers form their assumptions about what technology is for. She is worried those minds are being shaped by tools built for engagement and efficiency rather than curiosity and critical thought.

What does this mean for ordinary people?

The practical implication is less about which app you use and more about who you vote for and what questions you ask of the companies building these systems. Rundell's piece is a prompt, not a manual.

For parents and teachers, the more immediate question she raises is whether the version of AI students meet at school builds their thinking or quietly replaces it. That debate is happening in staffrooms and policy documents right now, and it will not wait.

Common questions

Is AI actually being used for surveillance already?

Yes. Governments in several countries already use AI-powered facial recognition and data analysis to monitor citizens. The technology is not hypothetical.

Can AI really help find a cure for cancer?

AI tools are already used in cancer research to analyse medical scans and identify drug candidates faster than human researchers alone. Results so far are promising but cures remain years away.

Should I be worried about AI and my child's education?

The debate is live and unsettled. The most useful thing a parent can do right now is ask their child's school what its policy on AI tools is and whether critical thinking is still explicitly taught.