Dragos: AI prompts leak planted secrets
Trackable secrets test · three providers breached
Security
Reuters cites internal posts; Meta has not responded
Written with AI assistance, reviewed by the ACN desk.
“This seems like an extreme privacy violation.”
Meta employees have raised concerns that its AI agent Muse mixes up conversations across accounts and exposes private user data. TechCrunch reports that the claims come from a Reuters report based on internal company posts.
Employees said Muse surfaced childcare requests they never made and career advice they never asked for. They also described product orders they never placed. Some saw bank account balances that did not belong to them. One employee said Muse sent bank balances unprompted during a chat about weather and tides.
Another employee, using the agent on a personal phone, got a reminder to answer a Signal message from her sister. “This seems like an extreme privacy violation,” the employee wrote. She said she had not granted the agent access to other apps.
Meta launched Muse in September. TechCrunch says it is one of the fastest-growing apps on record, with more than 10 million downloads. Meta did not immediately respond to TechCrunch’s request for comment.
“This seems like an extreme privacy violation.”
Sponsor
Spot an error? Email corrections@aicodingnews.si. Fixes are logged on the corrections page.
Sponsor
Sponsor
Trackable secrets test · three providers breached
Attackers used the open-source tool Cairn
A synthetic internet for prompt-injection tests
Poisoned agent-framework versions stole credentials
Sponsor
Sponsor