Latest in Gear

Image credit: ASSOCIATED PRESS

Facebook is relying on AI to tackle revenge porn

It aims to remove offensive images before a user reports them.
197 Shares
Share
Tweet
Share
Save
ASSOCIATED PRESS

Facebook is introducing new revenge porn detection tools that use AI to track down intimate images shared without the subject's consent. It claims machine learning can help it to detect near nude images and videos before a user reports them. A human moderator then reviews the flagged content, removing it if it violates Community Standards. Sharing revenge porn will almost certainly result in Facebook banning the account it originated from, said the company, though an appeals process is available in case of incorrect takedowns.

The tools will be rolled out to both Facebook and Instagram. They also include resources for victims -- such as contact information for local support groups and a guide on how to report content -- within Facebook's Safety Center. The tools are a result of user feedback including discussions with victims, advocates, and academics from the UK, Kenya, and Denmark. Facebook says it also tapped non-profits in the US, Pakistan, Brazil and South Korea to help create culturally and locally relevant resources for victims.

The new tools find Facebook trying to counter a wider range of material, including what Sheryl Sandberg described to People as "subtle abuses." But it hasn't given up on its confounding pilot program that asks potential victims to submit their own nudes. Facebook then creates a copy of the uploaded image to prevent it from being shared in the future. "We will expand this pilot over the coming months so more people can benefit from this option in an emergency," said Facebook in a blog post.

"Our research indicates we can improve our reporting tools to ensure they are easier to use and don't frustrate those who try," Facebook's Karuna Nain, global safety policy programs manager, and Radha Iyengar, head of product policy research, wrote in a blog post. "Victims we spoke with, for example, said they were put off by what they felt was Facebook's robotic response."

From around the web

ear iconeye icontext filevr