Images sent over X’s new encrypted chat service retain full EXIF metadata, including GPS coordinates, device information, and timestamps, Straight Arrow News (SAN) revealed this week. The finding, confirmed during a controlled test with early beta access, exposes a glaring privacy gap in a tool marketed as end-to-end encrypted. A single photo shared via XChat allowed SAN to identify a Google Pixel 8 Pro and pinpoint the shot to a Kansas City airport parking lot on May 12 at 10:45 a.m. This lapse is not an isolated oversight; it sits alongside a weak key-storage design that puts supposedly private conversations within reach of X itself.

The disclosure lands just as X rolls out the feature to a wider audience, aiming to bring secure messaging to its massive user base. But security experts and cryptographers are sounding alarms. Martin Shelton, deputy director of digital security at the Freedom of the Press Foundation, told SAN, “An end-to-end encrypted chat application should be architected in such a way that even the service provider, by design, cannot read the messages, even if they wanted to. Right now, it appears XChat does not provide these basic protections.”

The EXIF Oversight That Breaks Location Anonymity

EXIF (Exchangeable Image File Format) data is a bundle of metadata embedded in almost every photo taken with a modern smartphone or digital camera. It can include GPS coordinates, camera model, exposure settings, and the exact date and time a picture was captured. While professional tools and some social media platforms strip this information by default, XChat does not. SAN’s test proved that images sent through the encrypted channel arrive with all metadata intact—a practice that stands in sharp contrast to every major secure messaging app.

“Social media sites and chat apps have developed a good norm of stripping metadata by default, which prevents this sort of inadvertent disclosure,” said Jacob Hoffman-Andrews, senior staff technologist at the Electronic Frontier Foundation. “And people who use those apps have developed an expectation that metadata is stripped. If XChat is failing to strip metadata, it’s putting its users at risk.” That risk is not theoretical: a journalist sharing a photo from a sensitive meeting, a whistleblower documenting wrongdoing, or an activist coordinating a protest could all unintentionally broadcast their precise location and the device they use.

Even for everyday users, the threat is real. Stalkers, abusers, and doxers routinely mine EXIF data to track targets. A photo of a child’s first day at school, sent in an encrypted chat that lulls the user into a false sense of security, can reveal the school’s GPS coordinates and the time the family was there. Cybersecurity professional Ryan Montgomery summed up the reaction of many: he told SAN he was “absolutely mind blown” that XChat doesn’t strip EXIF data.

The technical fix is trivial. Apps like Signal automatically remove EXIF before uploading; iOS and Android APIs offer simple methods to scrub it. X’s failure to implement this one step suggests either a rushed development cycle or a fundamental misunderstanding of the privacy contract implied by “end-to-end encryption.”

Server-Held Encryption Keys and a Four-Digit PIN

While EXIF leakage is the most immediate and visible flaw, the deeper cryptographic design of XChat is raising even sharper concerns. To enable end-to-end encryption, each user needs a private key—a secret that only they should control. In mature systems like Signal’s, that key never leaves the device; the service provider has no way to decrypt messages, even under legal compulsion.

XChat takes a different path. According to X’s own documentation, the private key is encrypted with a four-digit PIN and then stored on X’s servers. The PIN acts as the sole protection of the key material. A four-digit PIN has only 10,000 possible combinations, making it trivial for anyone with server access to brute-force if they can get their hands on the encrypted blob. Worse, there is no rate-limiting inherent in such an attack when the attacker has the data off-server.

Matthew Green, a cryptography professor at Johns Hopkins University, has long warned about the dangers of server-side key storage combined with weak user passwords. He and others point out that without hardware security modules (HSMs) or a robust threshold-cryptography scheme, the server operator effectively holds the keys to the kingdom. X’s own support pages admit that “a malicious insider or X itself” could compromise an encrypted conversation—a statement that contradicts the basic guarantee of end-to-end encryption.

Industry best practice dictates that private keys remain on user devices and that the protocol provides perfect forward secrecy (PFS). In Signal’s double-ratchet algorithm, each message gets a unique ephemeral key, so that compromising one message does not unravel the entire history. XChat currently lacks PFS entirely. If an attacker cracks that single long-term private key—or obtains it from X’s servers—they can read every message ever sent, past and future. X says it is “working on” forward secrecy, but shipping an encrypted messenger without it in 2024 is a regression to pre-Signal days.

Metadata Retention: Who Talks to Whom, and When

End-to-end encryption usually protects message content, but a complete privacy solution also shields metadata: the who, when, and how of communication. Signal, for example, goes to great lengths to minimize metadata exposure, using sealed sender and other techniques. X takes an almost opposite approach. The company’s help pages explicitly note that while message content is encrypted, “associated metadata (e.g., recipient, creation time, etc.) is not.” Furthermore, if a public Post is shared inside an encrypted chat, X will have a record that the Post was shared.

This metadata harvest is valuable for platform operations, advertising, and legal compliance, but it gives X a detailed social graph of private interactions. Analysts warn that such logs can be subpoenaed, leaked, or exploited for surveillance. For users in authoritarian regimes, journalists, or anyone concerned about profiling, the retention of conversational metadata destroys the anonymity that encrypted messaging is supposed to provide.

The Closed-Source Problem

No third party can verify XChat’s claims because the source code remains closed. Signal, Wire, Element, and other trusted messengers publish their client and server code, inviting independent audits and community review. X has promised to release a whitepaper and open-source the implementation at some point, but those promises are not fulfilled yet. Relying on opaque, unaudited crypto is a gamble that security experts universally discourage. “Trust us” is not a security model.

Windows and Desktop User Implications

The risks are amplified for Windows users who frequently move between devices and rely on desktop clients. Windows does not automatically strip EXIF from images; in fact, many popular image editors and default photo viewers preserve all metadata. A user might take a screenshot or edit an image in Paint, then share it over XChat assuming it’s safe by virtue of the encryption lock icon. The reality is that the underlying file still contains revealing data. Additionally, Windows environments often sync files to cloud services, where metadata may be further exposed in backups or shares.

Enterprise users face additional compliance headaches. If employees use XChat for work conversations—perhaps because it’s convenient and “encrypted”—the organization may be leaking location footprints, device fingerprints, and communication logs that violate data protection regulations like GDPR or industry standards such as HIPAA.

How to Protect Yourself Right Now

Until X overhauls its architecture, security-minded individuals and IT administrators should treat XChat as an experimental feature, not a secure channel. Immediate steps include:

  • Do not use XChat for sensitive conversations. Stick with audited, open-source messengers like Signal, Wire, or Element.
  • Manually strip EXIF data from images before sharing. On Windows, right-click the file, choose Properties > Details > Remove Properties and Personal Information. Use batch tools like ExifTool for larger sets.
  • Disable location tagging in your phone’s camera settings to prevent GPS coordinates from being embedded at capture time.
  • Be mindful of what you share. Avoid sending photos of homes, license plates, unique landmarks, or interior spaces that could be identified.
  • For organizations, implement automated metadata-stripping on all outgoing content and educate staff about the risks of using unvetted communication tools.

What X Must Do to Recover Trust

If X wants XChat to be taken seriously, it must deliver swift, verifiable improvements:

  1. On-device key management. Private keys must never leave the user’s device. Any server-side component should be limited to public key distribution, authenticated by HSMs.
  2. Perfect forward secrecy. Deploy a double-ratchet mechanism so that each message has its own ephemeral key.
  3. Default EXIF stripping. Remove or sanitize all image metadata before upload. This is table-stakes for a privacy-respecting messenger.
  4. Transparency and audits. Open-source the entire codebase and commission independent audits from reputable firms. Publish the results without delay.
  5. Metadata minimization. Reduce the platform’s ability to collect and store conversational metadata. Offer an opt-in mode that uses sealed sender techniques.

Each of these steps is technically achievable and can be publicly verified. Until they are implemented, the security community will rightly view XChat as a privacy theater.

The Verdict

XChat’s launch is a case study in how not to introduce encryption. By neglecting basic metadata hygiene and choosing a server-centric key architecture, X has created a product that provides the appearance of security without its substance. The EXIF exposure is a straightforward bug—or perhaps a design oversight—that could have dire consequences for unwitting users. The deeper cryptographic weaknesses make the entire system brittle.

For now, the best advice is simple: don’t trust your private conversations to XChat. Use tools designed with the threat model in mind, not those that bolt encryption onto a surveillance business model. When even the company’s own fine print admits that an insider could read your messages, the “end-to-end” label means little. True privacy demands better engineering, and X has yet to deliver it.