EU Chat Control Returns: Why 'Default Scanning' Is the New Policy Temptation

The First-Round Vote: Chat Control Is Back

Heise reported that the European Parliament has approved the proposed Chat Control regulation in a first-round vote, sending it back to the center of EU digital policy. The news reached the top of Hacker News with 537 points and 237 comments, a level of attention that signals how much technical concern the proposal still carries. Chat Control would require platforms to scan private communications for child sexual abuse material, a goal nearly everyone supports. The problem is how it asks them to do it.

This is not the proposal's first appearance. Earlier versions stalled amid objections from privacy advocates, cryptographers, and some member states. Its return suggests the policy momentum has shifted from 'can we do this?' to 'we must do this by default.' That shift is the real story.

From Exception to Default: The Infrastructure Turn

The same week, another EU safety measure was drawing attention: mandatory driver-monitoring cameras in new cars. On the surface, Chat Control and in-car cameras have nothing to do with each other. One scans messages; the other watches eyelids. But both share a deeper design pattern: they turn exceptional surveillance into default infrastructure.

Traditional surveillance usually begins with a trigger: a crime, an accident, a suspect, a warrant. The new model is different. Systems collect continuously, classify in real time, and explain themselves later. Artificial intelligence makes this pattern cheap enough to apply at scale. When screening becomes inexpensive, the political temptation is to redefine scannable as should be scanned. Users then face more than the risk of a data breach; they face a society in which 'not being monitored' is treated as its own kind of risk.

• Event-driven: suspicion, incident, warrant, investigation.
• Default-driven: continuous collection, algorithmic classification, retrospective justification.

Why End-to-End Encryption Cannot Survive Client-Side Scanning

End-to-end encryption works because only the sender and the recipient hold the keys. The platform, the police, and any attacker who compromises the server see only unreadable data. To scan encrypted messages for illegal content, a service must either hold the keys itself or inspect the content before it is encrypted. The first option destroys confidentiality; the second is usually called client-side scanning.

Client-side scanning is often presented as a compromise that preserves encryption while catching abuse. It is not. It installs a classifier on the user's device that reads material before it ever reaches the cryptographic envelope. That is a backdoor, even if it is labeled a safety feature. Once the mechanism exists, its target list can expand: drugs, copyright, fraud, disinformation, protest coordination. It also introduces new attack surfaces, false positives, and a chilling effect on private speech.

Child safety is a serious objective. But the mechanism chosen to pursue it determines whether secure communication remains possible at all.

AI Makes Screening Cheaper, Not Automatically Proportional

The Hacker News discussion emphasized a risk that goes beyond the specific text of Chat Control: AI is making large-scale content screening so cheap that policymakers may start to treat it as routine infrastructure. Lower cost does not change the legal questions. Any interference with private communications must still be necessary, proportionate, and subject to independent oversight.

AI classifiers also scale their mistakes. A one-percent false-positive rate sounds acceptable in a lab; applied to billions of messages, it generates an avalanche of erroneous referrals. Classifiers can be gamed, drift over time, and reproduce training biases. If they are deployed by default, those errors become a baseline cost imposed on every user.

The institutional response should be stricter than the technical one: narrow legal scope, judicial authorization for non-routine inspection, public transparency reports, and sunset clauses that force legislators to justify renewal.

What Comes Next: Who Controls the Default?

The Parliament's first-round approval is not the final word. The Council and the trilogue negotiations still lie ahead. But the vote is a warning that the default is moving. The question is no longer whether platforms can detect abuse; it is whether detection should be the starting condition of every private conversation.

Technologists can help by designing detection systems that are auditable, constrained, and privacy-preserving. Regulators and citizens should ask the harder question: if scanning becomes opt-out, then privacy becomes opt-in. The history of digital infrastructure suggests that defaults are hard to undo. Once 'scannable by default' is built into law and code, the boundary between public safety and routine monitoring becomes very difficult to redraw.