The safety poison pill, narrowed

Victims have liberty too

Freedom is not maximized by refusing to prevent foreseeable victimization. Fraud, exploitation, credible threats, malware, and other serious harms can justify intervention. A cognitive-liberty framework that cannot admit this becomes absolutism rather than governance.

Not every safeguard is surveillance

A warning before a malicious site, a transaction check, a narrowly scoped classifier, or a request-level restriction can impose very different liberty costs from building a permanent dossier of everyone’s private inquiries. The architecture and scope matter.

The poison appears when safety becomes an open-ended jurisdiction

The historical concern is strongest when institutions collect broadly, define risk vaguely, retain data indefinitely, hide rules, prevent appeal, and cannot demonstrate marginal effectiveness. Those conditions create the infrastructure for secondary use and mission creep.

Least restrictive means is the practical bridge

Ask whether the same protective outcome can be achieved with a warning, user control, rate limit, sandbox, local classifier, anonymous credential, targeted investigation, or other narrower intervention before creating generalized surveillance.

Design against failure before the emergency

It is easier to define deletion, scope, oversight, and appeal before a crisis than after an institution becomes dependent on the new power. The lesson from past surveillance expansion is not that every future safeguard is illegitimate; it is that safeguards need safeguards.