Facebook launches a new Transparency Center, a site for info on its policies to tackle hate speech, says it now removes 97% of hate speech before it's reported
Facebook introduced its new Transparency Center Wednesday, billing it as a single destination for information on its integrity and transparency efforts.
AdweekDavid Cohen
Context & Ripple Effects
Facebook has been publishing quarterly enforcement statistics for years — proactive detection rose from 24% of hate speech takedowns in 2017 to 80% in late 2019 and 95% by its Q3 2020 transparency report. The new Transparency Center converts that rolling disclosure into a permanent, single-destination site for integrity policies and metrics.
It also follows an established playbook: the climate change information center launched in September 2020 showed Facebook packaging topical integrity work into dedicated hubs. What changes here is consolidation — one home for everything, headlined by a 97% pre-report removal rate that extends the same upward detection curve.
First-order effects
Facebook's headline metric — 97% of hate speech removed before anyone reports it — continues a trajectory it has disclosed each quarter, moving enforcement proof from scattered reports into one always-on destination.
Second-order effects
Rival platforms now face pressure to match both the metric format (proactive-detection percentage) and the infrastructure (a consolidated transparency hub), turning self-published enforcement data into a competitive benchmark.
Researchers and regulators get a single reference point for Facebook's claimed prevalence figures — which cuts both ways, since those numbers become the target for independent scrutiny.
Third-order effects
Self-reported prevalence becomes the contested terrain of platform accountability: when outside reporting challenged Facebook's hate speech numbers, the company countered with its own figure — prevalence down 50% over three years to roughly 0.05% of views — in its dispute over the WSJ's reporting, a fight made sharper by a hub that centralizes the official stats.
If the pattern holds, transparency infrastructure itself becomes a standing function of major platforms — published metrics and dedicated hubs as the default response to regulatory and press pressure rather than one-off disclosures.
The trend: Platform trust-and-safety is shifting from episodic transparency reports to permanently published, AI-measured enforcement metrics that companies use to frame — and defend — their own record.
Important detail: fb has hired EY to start doing independent audits of its transparency reporting Ofc, no one knows how such audits will work bc no one has ever done one before, but the industry is going to have to work it out bc it's desperately needed https://about.fb.com/...
Reminder: auditors' financial audits are considered important because they adhere to agreed standards and are regulated against them. Hiring an auditor firm to do whatever you ask is not the same thing https://twitter.com/...
Today we released our Community Standards Enforcement Report for Q1 2021: https://about.fb.com/.... As part of today, we're also launching a new Transparency Center, a single destination for all of our integrity and transparency efforts: https://transparency.fb.com/ (1/5)
For the second successive quarter @Facebook have taken down less than half the child abuse they did in Q3, because of successive technical issues. This is significant - but we're only finding out about it now https://transparency.fb.com/ ...
This is getting impressive results. Prevalence of hate speech, for example, has almost halved over the last year: https://about.fb.com/... https://twitter.com/...
Today, we are launching a new Transparency Center that we hope will give a clearer, more holistic view of our integrity enforcement work—and ultimately help external stakeholders more easily and effectively do their jobs (holding us to account!) → https://transparency.fb.com/ htt…
The prevalence of Hate speech on Facebook decreased for the second quarter in a row. In Q1, it was 0.05-0.06%, or between 5 and 6 views per 10,000 views. (4/5) https://twitter.com/...
On Q1's report: we added more prevalence metrics for IG - prevalence of adult nudity and of violent & graphic content. The prevalence of adult nudity on both FB and IG was 0.03-0.04%. Violent & graphic content prevalence was 0.03-0.04% on FB and 0.01-0.02% on IG. (3/5) https://tw…
NEW: In the first quarter of 2021, Facebook reported removing more organized hate content than terrorist content for the first time since these reports began in 2017. Could signal a post Jan. 6 shift in the trends I laid out here: https://www.protocol.com/... https://www.protocol…