Every message sent today probably passes through a server capable of reading it. Even apps marketed as "encrypted" often collect metadata: who you talked to, when, how often, from where. Metadata alone can be more revealing than the message itself.

In the last edition of Crypto 101, we looked at how AI is moving off the cloud and onto your device. This edition follows the same thread into communication: what actually happens when you hit send, why lawmakers keep circling back to your private messages, and how cryptography, not surveillance, is the real fix.

End-to-end encryption, or E2EE, scrambles a message on the sender's device using the recipient's public key. Only the recipient's private key, which never leaves their device, can unscramble it. The provider running the servers in between never holds the private key, so it structurally cannot read the content, even under legal pressure. Group chats make this harder. Multiple people need to share a key without ever being online at the same time, and that key needs to keep changing as members join or leave. This is exactly what the Messaging Layer Security (MLS) protocol solves. Standardized by the IETF as RFC 9420 in 2023, MLS provides efficient group key agreement for groups ranging from two people to thousands. Two properties matter most here. Forward secrecy means that if someone steals today's encryption key, they still cannot decrypt yesterday's messages, because keys constantly rotate. Post-compromise security means that if someone steals today's key, the system heals itself, and future messages become unreadable again once new key material is introduced.

This shift just went mainstream. In May 2026, Apple announced that end-to-end encrypted RCS messaging was rolling out in beta between iPhone and Android, built on the GSMA's RCS Universal Profile 3.0 and the same MLS standard. For the first time, a message sent from an iPhone to an Android device can be encrypted in a way that neither Apple, Google, nor the carrier can read.

Encryption protects what you said. It does almost nothing to protect who you said it to, when, and how often, and metadata alone can be more revealing than content. Metadata patterns from something as simple as call records have been shown to identify individuals with striking accuracy and to predict movement patterns. Three examples worth sitting with. A call to a crisis hotline at 2 a.m. lasting 45 minutes. The words are private, but the pattern alone tells the story. A journalist repeatedly contacting someone inside a government agency. The content is encrypted, but the contact pattern alone can expose a source. Messages to a divorce lawyer, then a realtor, then a moving company, all in the same week. No message needs to be read for the narrative to be obvious.

Different apps handle this very differently. WhatsApp collects extensive metadata, including contacts, device info, IP address, and group membership. Telegram stores plaintext messages server-side by default. Signal minimizes metadata through a feature called Sealed Sender, which hides sender identity from its own servers.

Two different proposals get conflated in most coverage, so it's worth separating them clearly. Chat Control 1.0 is a temporary, voluntary derogation from EU ePrivacy rules that lets platforms scan for known CSAM if they choose to. It expired on April 3, 2026 after MEPs initially voted 311 to 228 against extending it on March 26.

It was then revived on July 9, 2026 through a procedural mechanism. A motion to reject the extension needed an absolute majority of 360 votes to succeed, and only reached 314, falling well short despite more MEPs voting to scrap it than to keep it. That meant the extension passed by default rather than through a direct vote in its favor, as reported by The Register and Euronews. The revived rule now runs until April 3, 2028, and the amended text excludes traffic that is genuinely end-to-end encrypted, according to Brussels Signal.

Separately, and more consequentially, Chat Control 2.0, the permanent CSA Regulation, remains stuck in trilogue negotiation. Five rounds of talks, including what was meant to be a final session on June 29, have passed without agreement on its shape, and it could still impose mandatory scanning obligations that critics argue are far harder to reconcile with encryption.

Age verification and KYC-style identity checks are spreading across platforms in parallel with scanning debates. Each check creates a new centralized linkage: a real name or government ID tied permanently to an online account, and to every message sent from it. Layer a scanning regime on top of an identity mandate, and anonymous, private communication effectively disappears. What's left is a searchable, attributable record.

The real danger isn't one bad law. It's what happens when a scanning mandate, an ID mandate, and a single breached database all line up at once. Content, identity, and metadata, all linked, all in one place.

This is where the story pivots from problem to solution, using XMTP as the flagship example. As the XMTP team put it when outlining their mission: privacy is under attack, and centralized messaging services keep giving in by surrendering user data or discontinuing privacy features. XMTP is built to be a different kind of network entirely: no central server capable of reading messages, and no single company or country as a point of failure, as laid out in their decentralization announcement. XMTP's identity model centers on an Inbox ID, a stable destination for messages that stays constant even as a user adds or removes underlying identifiers. A user can link multiple identity types to one inbox: Ethereum wallets, smart contract wallets, passkeys, World ID for proof of personhood, and social identifiers like ENS or Farcaster. Every link is a cryptographically signed action, verifiable by anyone on the network without needing to trust a central authority to vouch for it.

Each inbox also has a designated recovery identity, a cryptographic emergency key that can revoke a compromised identifier or rotate control. This prevents the classic "lost my phone number, lost my account" failure mode of centralized platforms. XMTP's own decentralization progress update confirms the network is quantum-resistant end to end, inheriting forward secrecy and post-compromise security from MLS, so even encrypted traffic recorded today stays protected against tomorrow's more powerful computers.

This is the answer to the false choice. Chat Control-style laws frame the tradeoff as privacy versus child safety, or anonymity versus accountability. Decentralized identity shows a third path: prove a specific fact, like "I am a unique human," without revealing your full identity to the platform, the government, or advertisers. Proof without exposure is the whole point.

Encryption protects what you say. The next piece of this puzzle is protecting who you are while you say it. That's where E34 picks up.