Privacy

Meta Pauses AI Tracking Program After Data Leak

Is this a reckless breach of trust or a responsible foundation for building better AI?

JUN 24·Updated JUN 29

Meta Pauses AI Tracking Program After Data Leak

Story

Above: The Meta logo at the Mobile World Congress 2026 (MWC) at the Fira de Barcelona, Spain, on March 2. Image credit: Davide Bonaldo/SOPA Images/LightRocket/Getty Images

The Facts

U.S. tech giant Meta paused its Model Capability Initiative (MCI), an internal AI training program, on Monday after realizing that substantial amounts of employee data collected by the MCI had been left accessible to all staff.
BBC News
TNW
Documents seen by Business Insider and WIRED revealed that the leaked information included private conversations, complete transcripts, performance and prompt data, and that Meta had classified the incident as a SEV 2 on its 0-to-5 scale, with 0 being the most severe.
Business Insider
Wired
In a statement to multiple media outlets, Meta stated that MCI was built with "carefully designed" privacy safeguards and that it had found "no indication at this time that any data was improperly accessed." The company added that it would nevertheless pause the program while it investigated.
PCMAG
Fast Company
Meta CTO Andrew Bosworth, however, acknowledged to employees in an internal memo obtained by WIRED that MCI's implementation had fallen short of its privacy review standards, citing "misconfigured ACLs" as the cause of the exposure.
Wired
The company originally launched the program in April, which logged the mouse movements, keystrokes and navigation behavior of U.S.-based employees on their work devices, supplemented with occasional screenshots, to generate training data for Meta's AI systems.
Mashable
FOX News
MCI quickly drew backlash, with more than 1,600 employees signing a petition opposing the program, which warned it posed "security and regulatory risks for Meta." In response, Meta announced earlier this month it would allow employees to opt out of the tracker, but only for 30 minutes at a time.
Independent
Reuters
Wired

Sources Split

LEFT

CENTER

RIGHT

ESTAB
CRIT

CENTER

PRO
ESTAB

The Spin

Establishment-critical narrative

Meta's MCI program was a disaster waiting to happen, with this invasion of privacy leaving everything from employees' keystrokes to their private conversations exposed for anyone inside the company to view. Most damning of all, despite promising tight controls, Meta has evidently failed to deliver, signaling a fundamental breach of trust between the company and its workforce.

Futurism

Pro-establishment narrative

Meta's AI development relies on responsibly sourced data, processed with extensive privacy safeguards such as de-identification and aggregation. The goal is building AI that's accurate, safe and genuinely useful for everyday tasks. Continuous improvement in this way is exactly how responsible AI gets built, and that process deserves recognition rather than reflexive suspicion.