Are tech companies removing evidence of war crimes?

2 years ago 41
ARTICLE AD BOX

By James Clayton
North America technology reporter

Image source, Mike Kemp/Getty Images

In just three months, last year, TikTok removed 80 million uploaded videos that in some way broke its rules.

Powerful artificial intelligence combined with human moderators had removed them at lightning speed, 94.2% before anyone had seen them, it said.

Systems that searched for "violative content" had removed 17 million of the videos removed automatically.

And other social-media companies share a similar story - thousands of hours of content taken down every day.

Now, some are asking whether Big Tech, by removing so much content, is also removing footage of war crimes in Ukraine.

Graphic content

TikTok was already hugely popular around the world before Russian President Vladimir Putin's decision to invade Ukraine - but the war has been a coming-of-age moment for the platform.

Videos using various Ukrainian hashtags have had billions of views.

But Ukrainians uploading videos from the ground could be generating more than "likes".

They may well be uploading a piece in a jigsaw of evidence that will one day be used to prosecute war crimes.

But they may also be breaking TikTok's and other social-media companies' strict rules on graphic content.

"TikTok is a platform that celebrates creativity but not shock-value or violence," TikTok's rules say.

"We do not allow content that is gratuitously shocking, graphic, sadistic or gruesome."

And some, but not all, content depicting possible atrocities could fall into that category.

'Huge issue'

Researchers are unclear how much Ukrainian user-generated content Tiktok, and other social-media companies such as Meta, Twitter and YouTube are taking down.

"TikTok is not as transparent as some of the other companies - and none of them are very transparent," Witness programme director Sam Gregory says.

"You don't know what was not visible and taken down because it was graphic, but potentially evidence.

"There's a huge issue here."

Image source, Maxar Technologies

Image caption,

A satellite image showing the aftermath of the air strike on the Donetsk Academic Regional Drama Theatre, in Mariupol, Ukraine

This is not the first time big social-media companies have had to deal with evidence of potential war crimes.

The Syria conflict threw up similar problems.

Human Rights Watch has called for a centralised system of uploads from conflict zones for years, without success.

"At the moment, it doesn't exist," senior conflict researcher Belkis Wille says.

She describes the haphazard and convoluted process prosecutors have to follow to obtain evidence removed from social media.

"Authorities can write to the social-media companies, or ask for a subpoena or court order…. but the way the process works right now, no-one has a very good picture of where all this content is," Ms Wille says.

And that is a real problem for investigators.

Even before Ukraine, those trying to document atrocities highlighted how increased moderation was having a detrimental effect on evidence gathering.

Image source, Getty Images

"This pace of detection means that human-rights actors are increasingly losing the race to identify and preserve information," a report into digital evidence of atrocities, by the Human Rights Center, at Berkeley School of Law, said.

The report called for "digital lockers" - places where content can be stored and reviewed not just by social-media companies but by non-governmental organisations (NGOs) and legal experts.

But many social-media companies do not want to invite outsiders into their moderation processes, leaving researchers with a Rumsfeldian conundrum - they often do not know what has being taken down, so how can they know what to request or subpoena?

These are unknown unknowns.

Light-touch policy

But not all social-media platforms have the same policies when it comes to graphic content.

Telegram has been hugely important in sharing videos from the ground in Ukraine.

It also happens to have an extremely light-touch policy on moderation.

Videos that would be taken down on Twitter or Facebook stay up on Telegram.

And that is not the only reason the platform is helping investigators.

"I would say some of the most valuable photo and video content that we as an organisation have received is from Telegram," Ms Wille says.

And there is another key benefit.

Social-media companies such as Facebook and Twitter automatically strip content of a picture or video's metadata - a kind of digital ID, revealing where and when the content was captured and crucial for investigators.

"One benefit we've found is that the metadata is not stripped on Telegram," Ms Wille says.

Protecting the metadata, from the moment an action was captured to the moment it is shown in court, is sometimes referred to as "chain of custody".

Wendy Betts, the director of Eye Witness, a project of the International Bar Association focused on the collection of verifiable human-rights atrocities, encourages people to film possible war crimes on its app, Eye Witness to Atrocities, to preserve the information in the best possible way for use in court.

"As footage passes from photographer to an investigator to a lawyer… if any link in that chain is missing, that footage is going to be looked at as more suspect, because changes could have been made during that gap," she says.

But all these solutions feel piecemeal and unsatisfactory.

With no one digital locker all social-media companies use, with no one place where all this is being stored, crucial evidence could be falling down the cracks.

Different responses

In some cases, it is not clear social-media companies are storing or documenting these videos at all.

BBC News asked TikTok, Google, Meta and Twitter about their policies in this area.

TikTok forwarded its policies on protecting its users during the Ukraine war but failed to address any of the questions asked.

"We don't have more to share beyond this information right now," a representative said.

And neither Twitter nor Google responded.

Only Meta gave a tailored response.

"We will only remove this type of content when it glorifies the violence or celebrates the suffering of others or when the content is extremely graphic or violent - for example, video footage of dismemberment," a representative said.

"In relation specifically to the war in Ukraine, we are exploring ways to preserve this type and other types of content when we remove it."

The very different responses from these four huge technology companies tells its own story.

There is no system, no policy, they all share.

And until there is, crucial evidence could be being lost and forgotten.

War in Ukraine: More coverage

Read Entire Article