Chat Control, Explained: The EU's Fight Over Scanning Your Messages

EU Chat Control is the activist nickname for the EU’s proposed Child Sexual Abuse Regulation (the CSA Regulation, or CSAR) — a draft law that, in its original 2022 form, would have forced messaging apps, email providers, and hosting services to scan everyone’s private messages, photos, and files for illegal content. As of mid-2026 it is not law: the most controversial part, mandatory scanning, was dropped from the Council’s text in late 2025, a separate interim voluntary-scanning law lapsed in early April 2026, and the permanent regulation is still in closed-door negotiation. The fight isn’t over — and the goal it claims to serve, protecting children, is genuinely important. The problem has always been the method.

This post explains what Chat Control actually proposes, why cryptographers say “client-side scanning” breaks end-to-end encryption even when the encryption technically stays in place, how the political fight unfolded through 2025 and 2026, and where it stands now.

What is EU Chat Control, in plain terms?

Imagine the post office promised never to open your sealed letters — but then installed a machine in your own home that photographs every letter and compares it against a watchlist before you seal the envelope. The letter still travels sealed. But something already looked inside. That is the core idea behind Chat Control, and it is why the “your messages are still encrypted” reassurance rings hollow to security researchers.

In precise terms: the CSA Regulation (“CSAR,” sometimes called “Chat Control 2.0”) was proposed by the European Commission in May 2022. Its stated aim is to make online services detect, report, and remove child sexual abuse material (CSAM) and to detect grooming. The controversial enforcement tool was the detection order — a legal command that could compel a provider to scan its users’ communications for three categories of content:

• Known CSAM — matched against a database of hashes (digital fingerprints) of previously identified images.
• Unknown CSAM — flagged by AI classifiers trained to recognize new abuse imagery they have never seen.
• Grooming — detected by text-analysis models reading the actual words in conversations.

The first category is the narrowest and most accurate. The second and third are where the trouble multiplies: AI image classifiers and text-based grooming detectors produce far higher false-positive rates, meaning innocent family photos, medical images, or ordinary conversations can get flagged and forwarded to human reviewers or authorities. The original mandate covered all three — a fact often lost when supporters describe it as merely “checking for known illegal images.”

Does Chat Control break end-to-end encryption?

Yes, in the view of essentially every independent cryptographer — and understanding why requires looking past the political talking point.

End-to-end encryption (E2EE) means only the sender and the intended recipient can read a message; the service carrying it, and anyone tapping the wire, sees only ciphertext. Apps like Signal and WhatsApp, and architectures like RVNT’s, are built so that no middle party — not even the app maker — holds the keys.

For an E2EE app, there is only one way to comply with a scanning order: client-side scanning (CSS). Because the provider cannot read encrypted messages in transit, the scanning has to happen on your device, before encryption is applied (or after decryption on the receiving end). The content gets inspected at the one place it briefly exists in the clear — inside your own phone.

Defenders argue this “preserves encryption” because the message is still encrypted on the wire. Cryptographers reject that framing as a category error. The guarantee E2EE makes is not “the bytes are scrambled between two points.” The guarantee is confidentiality: that no one but the participants learns the content. A mandated scanner sitting inside the endpoint, reporting matches to a third party, is that someone. The encryption becomes theater around a hole already cut in the wall.

In 2021, a group of prominent security researchers — including Hal Abelson, Ross Anderson, Whitfield Diffie, Ronald Rivest, and Bruce Schneier — published “Bugs in Our Pockets: The Risks of Client-Side Scanning” in the Journal of Cybersecurity. Their conclusion was blunt: CSS creates serious risks for security and privacy, cannot be limited to its stated purpose, and builds infrastructure that is one policy change away from scanning for anything. You cannot, they argue, construct a back door that only “the good guys” can use — once the scanning hook exists on the device, the only questions left are what it looks for and who gets to decide. This is the textbook definition of mission creep: a tool built for child protection becomes available for copyright, “extremism,” dissent, or whatever the next government defines as in scope.

The Internet Society’s technical analysis reaches the same place from an infrastructure angle — CSS turns every user’s device into a surveillance checkpoint and weakens the security of the entire system, not just the targeted content.

How does this compare to Apple’s abandoned 2021 CSAM plan?

Anyone who followed tech in 2021 has seen this movie. Apple announced a plan to scan photos on iPhones against a CSAM hash database before they were uploaded to iCloud — a real-world client-side scanning system from one of the most security-focused companies on earth.

Within weeks, researchers demonstrated hash collisions: they engineered innocuous images that produced the same NeuralHash as flagged material, showing the system could be fooled into false matches — and, more worryingly, that the hash list itself was an opaque, repurposable lever. Apple paused the plan and later abandoned it entirely. The same technical objections that ended Apple’s voluntary, single-vendor system apply with greater force to a mandatory, EU-wide, multi-provider version. The math does not care who is ordering the scan.

Chat Control 1.0 vs Chat Control 2.0: what’s the difference?

These two are constantly confused, so it is worth separating them cleanly:

	“Chat Control 1.0"	"Chat Control 2.0”
Formal name	Regulation (EU) 2021/1232	CSA Regulation (CSAR)
What it is	Temporary derogation from the ePrivacy Directive	Permanent proposed regulation
What it does	Permits providers to voluntarily scan non-encrypted messages	Would have mandated detection (incl. via client-side scanning)
Scope	Gmail, Messenger, Instagram, Outlook, Xbox, etc.	All messaging, email, hosting — including E2EE apps
Status (mid-2026)	Lapsed in early April 2026	Still in trilogue negotiation

Chat Control 1.0 was the stopgap that allowed (did not require) big platforms to scan unencrypted messages. Chat Control 2.0 is the permanent, far broader law meant to replace it — and the one that introduced the mandatory-scanning detection orders.

Is Chat Control law yet? The 2026 status

No. As of mid-June 2026, no final text has been adopted, and it would be wrong to say Chat Control “passed” — just as it would be wrong to declare it dead. Here is how the fight actually unfolded:

• May 2022 — The Commission proposes the original CSAR with mandatory detection orders.
• November 2023 — The European Parliament adopts its negotiating position, with its lead committee rejecting indiscriminate scanning and the breaking of encryption. Parliament would allow detection only under a targeted, judicially authorized order against specific suspects or groups.
• 2022–2025 — The Council (the member-state governments) deadlocks repeatedly, unable to reach a qualified majority on mandatory scanning.
• September–October 2025 — The Danish Council Presidency revives a text that still allowed mandatory scanning, triggering renewed public outcry.
• Early October 2025 — Germany signals it will not support mandatory scanning, joining a blocking minority alongside countries including Austria, Poland, the Netherlands, Finland, the Czech Republic, Luxembourg, and Slovenia — enough to stop the proposal.
• October 2025 — The scheduled Council vote on the mandatory-scanning text is pulled.
• Late October 2025 — Denmark announces it will drop the mandatory detection-order / mandatory client-side-scanning requirement, shifting to a voluntary model.
• November 26, 2025 — The Council (via its Permanent Representatives Committee) endorses the revised text as a “general approach” to take into trilogue — a close, divided vote. Critics, including MEP Patrick Breyer, warned that broad “risk-mitigation” language could reintroduce scanning through the back door.
• December 2025 — Trilogue (the closed-door Parliament–Council–Commission negotiations) on the permanent regulation gets underway; separately, the Commission seeks to extend the interim derogation to keep voluntary scanning alive.
• March 26, 2026 — The European Parliament votes 311–228 (with abstentions) against extending Chat Control 1.0.
• Early April 2026 — The Chat Control 1.0 derogation lapses (sources cite April 3 or April 4). Voluntary scanning of unencrypted messages is left without a clear EU legal basis — though some providers indicated they would continue scanning, and enforcement of the lapse is unclear.
• April–June 2026 — Further trilogues on the permanent regulation (sessions reported around mid-April and May), now focused on detection rules, encryption, and age verification, with a session scheduled for late June.
• Mid-2026 onward — If a deal is reached, formal adoption by Parliament and Council could follow, with some trackers pointing to a possible July 2026 target. As of this writing no deal exists, and the timing could slip.

One honest caveat on the record: reporting on the exact March 2026 vote is not perfectly uniform. The most widely cited account — corroborated by EDRi, CDT Europe, and multiple trackers — is that Parliament rejected the extension 311–228 and the derogation lapsed. A minority of outlets framed the same period as a vote to extend the derogation with new limits. We follow the better-corroborated rejection account, and flag the divergence rather than paper over it.

Would Chat Control affect Signal and WhatsApp — would they leave the EU?

Signal has been unambiguous. Its president, Meredith Whittaker, has said repeatedly that Signal would leave the EU rather than implement client-side scanning or otherwise break its encryption. Other privacy-focused providers, including the German email service Tuta, have echoed the same line: there is no version of “scan everyone’s messages” that is compatible with offering real end-to-end encryption, so a mandate to scan is effectively a mandate to stop offering the product.

This is the quiet leverage that helped stall the mandatory version. A law that drives the most trusted secure-messaging apps out of Europe — leaving residents with less secure tools — is a hard outcome for any government to defend, especially when those same governments rely on those apps for their own sensitive communications.

Is mandatory age verification part of Chat Control now?

This is one of the most important shifts, and the one most people have missed. With mandatory scanning largely off the table, much of the live battleground has moved to mandatory age verification — and digital-rights groups now treat it as a major near-term threat.

Age verification (or “age assurance”) means proving how old you are before you can use a messaging app: uploading a government ID, submitting to a face scan, or having your behavior analyzed to estimate your age. The privacy problems are severe:

• It can end anonymous communication. Tying a real identity (or a biometric) to every account erodes the ability to speak without being identified — a protection that journalists, whistleblowers, abuse survivors, and ordinary people all rely on.
• It creates new honeypots. Databases of IDs and face scans are exactly the high-value targets attackers love.
• It is a mission-creep vector. Infrastructure built to check ages can be repurposed to check identities for other reasons.

Both EDRi and the EFF have warned that mandatory age verification, layered with app-store age restrictions, could do lasting damage to private and anonymous communication even if not a single message is ever “scanned.” Civil-society groups have repeatedly described the broader effort as a proposal that keeps coming back from the dead and must not be allowed to return through the back door.

What can people do, and who is opposing it?

The opposition is broad and well-organized. The German digital-rights campaigner and former MEP Patrick Breyer maintains one of the most-cited running trackers of the proposal. EDRi (European Digital Rights) coordinates civil-society analysis and a canonical document pool. The EFF, CDT Europe, and national digital-rights groups have filed analyses at each stage. Their consistent ask of EU residents: contact your MEP and your national government’s representatives in the Council, because the proposal lives or dies on whether the blocking minority holds and whether Parliament keeps its “targeted, judicially authorized only” line in trilogue.

The architectural takeaway

Chat Control is the clearest real-world demonstration of why message security has to be designed in, not promised. Over four years the political winds reversed several times — mandatory, then voluntary, then expired, then back in trilogue, now pivoting toward age verification. Any system whose privacy depends on a vendor’s promise or a regulator’s restraint is one vote away from losing it.

The only durable protection is an architecture that structurally cannot scan: no server-side plaintext to inspect, no central party holding keys, and no client-side-scanning hook to mandate. If a system is built so that there is nothing to hand over and nothing to scan, the debate over detection orders becomes moot — not because the law was kind, but because there is no lever to pull. That is the design principle behind genuinely end-to-end-encrypted, peer-to-peer systems, and it is the reason the threat model matters more than any feature list.

To be clear about the limits — because honesty is the point — no architecture protects against a compromised device with malware on it, a contact who screenshots and forwards your message, or a legal order served on the person you are talking to. Encryption protects the channel, not the people on either end. But the Chat Control fight was never really about those edge cases. It was about whether the channel itself stays private by design. For now, the answer in Europe is still being negotiated. Do not assume it is settled — and do not assume it cannot come back.