Policies that outline what is and isn't allowed on the Facebook app.
Policies for ad content and business assets.
Other policies that apply to Meta technologies.
How we update our policies, measure results, work with others, and more.
How we help prevent interference, empower people to vote and more.
How we work with independent fact-checkers, and more, to identify and take action on misinformation.
How we assess content for newsworthiness.
How we reduce problematic content in News Feed.
Quarterly report on how well we're doing at enforcing our policies on the Facebook app and Instagram.
Report on how well we're helping people protect their intellectual property.
Report on government request for people's data.
Report on when we restrict content that's reported to us as violating local law.
Report on intentional internet restrictions that limit people's ability to access the internet.
Quarterly report on what people see on Facebook, including the content that receives the widest distribution during the quarter.
Download current and past regulatory reports for Facebook and Instagram.
This metric shows the percentage of all content or accounts acted on that we found and flagged before users reported them to us. We use this metric as an indicator of how effectively we detect violations.
Our investments in machine learning technology are critical to help us detect faster.
We balance machine learning with a trained team of experts who review and take action on violating content.
The rate at which we can proactively detect potentially violating content is high for some violations, meaning we find and flag most content before users do. This is especially true where we’ve been able to build machine learning technology that automatically identifies content that might violate our standards.
Such technology is very promising but is still years away from being effective for all kinds of violations. For example, there are still limitations in the ability to understand context and nuance, especially for text-based content. This creates additional challenges for proactively detecting certain violations.
The metric can go up or down due to external factors. For example, a cyberattack during which spammers share 10 million posts featuring the same malicious URL. If we detected the malicious URL before any user reported it to us, proactive rate would go up during the cyberattack and go down afterward, even if our detection technology didn’t change during the period. This metric can also increase or decrease based on how our processes and tools change — for example, it might go up if our detection technology gets better, but it might go down if our user reporting improves and we rely less on proactive detection.
Since this metric is based on the amount of content actioned, many of the same considerations apply. Our proactive rate doesn’t reflect how long it takes us to detect violating content or how many times it was viewed before detection. It also doesn’t reflect how many violations we failed to detect altogether or how many times that content was viewed. And though the percentage of content we proactively detect can be very high — as high as 99% in some categories — even the remaining small percentage can cause significant impact to people.
We calculate this percentage as: the number of pieces of content acted on that we found and flagged before people using Facebook or Instagram reported them, divided by the total number of pieces of content we took action on.
For fake accounts on Facebook, we calculate this metric as the percentage of Facebook accounts disabled for being fake that we found and flagged before users reported them to us. It’s calculated as the number of disabled accounts we found and flagged before users reported them, divided by the total number of accounts disabled for being fake.
We compute our proactive rate using a strict attribution of user reports to content. For example, if someone reports a Page and, while reviewing the Page, we identify and act on some violating content within that Page, we would report that we had flagged that content proactively (unless there were specific additional user reports of it). This strict way of attributing user reports is not ideal, but we haven’t yet determined a better method.