The Syllabus Is the Battleground

The pattern is old: when power narrows the terms of debate, the people who ask hard questions pack their bags and go where they can finish a sentence without a donor clearing a throat. Call it migration by necessity, not romance. The stakes aren’t academic. Whoever sets the syllabus for “AI security” and “cyber policy” decides whether the public gets warning labels or sales brochures. don’t trust the coffee.

Companies moved first. They built the systems, set the pace, and wrote the talking points. Everyone else is reacting.

I’ve watched security people do this shuffle for decades. When you can’t publish without a PR minder, you go where publication is still treated as oxygen. A season in Toronto isn’t about scenery. It’s about choosing an institution that lets you name names and release findings on a calendar, not at a vendor’s pleasure. A fall reading group on AI security sounds modest; it isn’t. Reading lists become agendas. What gets read gets regulated; what gets skipped becomes the loophole everyone calls too technical to fix. A spring class on cybersecurity policy matters for the same reason: you’re training the staff who will either cut through vendor fog or repeat it.

There’s another draw: muscle memory. The Citizen Lab earned it by proving mercenary spyware on journalists’ phones while officials elsewhere fussed over optics. Add the law school and the Schwartz Reisman Institute, and you get three legs under the same table: technical proof, legal framing, and governance work in the same room. That mix is rare at a time when Washington is rehiring industry to police itself and treating disclosure as a breach of etiquette.

The sequence is blunt. Firms sprint ahead with systems that classify, rank, and quietly punish. Government rules limp behind, often drafted by the firms’ alumni. That imbalance isn’t random. Money prefers an “AI safety” that means better prompt filters and less talk about workplace monitoring, invasive scoring, or police tools that misidentify people and then make them prove they aren’t the file. Universities promise ethics centers while licensing exam software that flags accents as “suspicious.” City halls sign cloud deals whose nondisclosure clauses bury data requests. Border officers still treat phones like suitcases and help themselves.

Start with first principles: the law must treat people equally and give each of us a real chance to challenge a machine’s decision before it hardens into a record. Here’s what it looks like when that standard is ignored. A benefits office denies aid because software can’t parse a worker’s split shifts. A résumé gets filtered out by a model trained on last year’s hires. A faulty match triggers a late-night visit from police and a name that never quite leaves the database. None of this is theoretical; it is routine, and it can land on anyone.

Timing tells you what matters. Corporations can ship a model in a quarter. Legislatures take years. So the classroom becomes the pressure point. A reading group that faces real threat models, surveillance landlords, automated discipline on the job, cross‑border device searches, can inoculate a generation against corporate euphemism. Duck those realities and lean on science fiction and safety pageantry, and you’re not educating; you’re training compliance.

Follow the incentives and the picture sharpens. The platform wants your faceprint and your work product, sold as one‑click convenience. The defense contractor wants a grant pipeline that stamps “dual‑use” on everything and moves debate behind closed doors. The consulting shop wants another deck to explain delay. The public pays in cuts you feel but rarely see coming: fewer hours, higher quotas, errors you can’t contest, data you can’t see but that still decides what you get.

People don’t relocate for novelty. They move so the work can be checked, argued, published, and used. If that space shrinks here, it will grow somewhere else. This is not a matter of mood; it’s a matter of whether evidence can be made public without permission from a sponsor. Institutions that still sign off on plain statements of fact—what a tool does, who it harms, how often, will inherit the work, and the trust.

Be clear about agency. Firms rewrote the pace and the terms. Academics and a handful of civil servants are scrambling to contain the fallout. If you want the next cohort to do more than rebrand surveillance as “safety,” build them rooms where technical findings, legal obligations, and policy levers are forced to sit together. Give outside reviewers authority to demand records, and penalties when records are withheld. Require impact notes before deployment, not after a headline. Publish error rates and appeal rights next to the product claims. Treat every automated denial as contestable, with timelines that favor the person, not the vendor.

None of this is heroic. It’s plumbing and paperwork. But it draws a boundary the market won’t draw on its own. If we leave the syllabus to the sellers, we get courses in how to live with the product. If we give it to labs with receipts, we get a record the public can use.

The choice is already on the table: who drafts the next rules, people who prove what systems actually do, or companies holding the contract? Choose carefully. The decision will set the tone for years, long after the press release turns to dust.

© Elias Rourke, 2025. Use it if it helps. Credit if you’ve got a conscience.