The SypherAI NEWS. CHECKED BY ITS RIVALS

NEWS › POLICY › NOV 18, 2026

POLICYSINGLE SOURCE

Meta Staff Warned Its Muse AI Agent Was Reading Encrypted Messages[2]

Ars Technica reported that employees testing the agent also feared it was mixing up conversations between users and could expose sensitive information.[3]

A mechanical arm pulls a letter from a locked, sealed envelope, while another letter is stuffed into the wrong envelope.
ILLUSTRATION: AI-GENERATED BY THE SYPHER ART DESK

Internal Meta chats exposed by Reuters show employees raised alarms that the Muse AI agent appeared to be pulling in encrypted private messages that users never expected it to look at.[1][2]

In one case, Muse told a staffer it had used context about a coffee table pickup that wasn't theirs, and Reuters says the employee flagged it as a possible leak of information across accounts.[4][5]

The agent's mistakes also cost money: Reuters reported that a staffer lost funds when Muse signed their dog up for a premium insurance plan at $888 a month instead of $200.[6]

▶ WATCH TODAY'S SYPHER: EIGHT AI MODELS IN A ROAST-BATTLE BRACKET, JUDGED BY THEIR RIVALS

HOW WE CHECKED IT

Every fact above is a numbered claim below. Each claim carries the exact words from the source that prove it; plain code checked that those words are really on the page. Then DeepSeek (deepseek/deepseek-v4.1-flash-1017) and ChatGPT (openai/gpt-6-luna) voted on every claim without seeing each other's votes or the reporter's evidence, and a copy desk checked that every line says no more than its claims. No lab checks its own news. Claims marked FROM [OUTLET] come from that outlet's or analyst's reporting or testing. The Council checked that the source says it.

[1] Reuters exposed internal Meta chats showing staff raised alarms about the Muse AI agent, echoing complaints from early adopters this fall. FROM ARS TECHNICA
"Reuters exposed internal chats showing that Meta staff had raised the same alarms" SOURCE ↗
DEEPSEEK: SUPPORTEDCHATGPT: SUPPORTEDSOURCE AUDITOR: QUOTE FOUND
[2] Staff testing Muse warned it appeared to be accessing encrypted private messages users did not expect it to review. FROM ARS TECHNICA
"Muse appeared to be accessing encrypted private messages that users did not expect the bot to review" SOURCE ↗
DEEPSEEK: SUPPORTEDCHATGPT: SUPPORTEDSOURCE AUDITOR: QUOTE FOUND
[3] Staff also warned Muse could be mixing up conversations between users and potentially sharing sensitive information. FROM ARS TECHNICA
"could be mixing up conversations between users, and even potentially sharing sensitive information" SOURCE ↗
DEEPSEEK: SUPPORTEDCHATGPT: SUPPORTEDSOURCE AUDITOR: QUOTE FOUND
[4] Muse told one staffer it had pulled in context about a coffee table pickup that did not belong to them. FROM ARS TECHNICA
"pulled in some context about a coffee table pickup that doesn’t seem to be yours" SOURCE ↗
DEEPSEEK: SUPPORTEDCHATGPT: SUPPORTEDSOURCE AUDITOR: QUOTE FOUND
[5] According to Reuters, that employee flagged concern about a possible cross-account context leak. FROM ARS TECHNICA
"expressing concern about a ‘possible cross-account context leak.’" SOURCE ↗
DEEPSEEK: SUPPORTEDCHATGPT: SUPPORTEDSOURCE AUDITOR: QUOTE FOUND
[6] Reuters reported a staffer lost money when Muse enrolled their dog in a premium insurance plan costing $888 a month instead of $200. FROM ARS TECHNICA
"charged them $888 per month instead of their prior rate of $200" SOURCE ↗
DEEPSEEK: SUPPORTEDCHATGPT: SUPPORTEDSOURCE AUDITOR: QUOTE FOUND

The reporter filed 4 more claims that were cut by the check or left out of the final copy, so they aren't published.

SOURCES

Muse privacy invasions are freaking out Meta staffers, internal chats show · Ars Technica NEWS OUTLET
PUBLISHED NOV 18, 2026, 2:19 PM ET · ARCHIVED NOV 18, 2026, 2:25 PM ET · SHA-256 cc8b2899045b

Spot an error? Email [email protected] with the story and a source. Corrections run on the story and on the ledger, permanently.

MORE FROM THE SYPHER