'Violating and dehumanising': How AI deepfakes are being used to target women

Experts describe the use of a person’s images to create deepfakes using artificial intelligence (AI) without their consent as "dehumanising".

Transcript

00:00 So my journey with image based abuse has actually gone on for around 10 years and it started

00:06 off with anonymous perpetrators taking images of me from social media and doctoring them

00:12 into pornographic images and over time this escalated in how severe it got, how graphic

00:21 the images were, how many images there were until it turned into deepfake videos.

00:28 You've probably heard of deepfakes. To put it simply, a deepfake is a video or an audio

00:31 file which has been altered using artificial intelligence. You can use it to reproduce

00:36 someone's voice or to swap faces on a video. The result is fake but comes over as extremely

00:41 realistic. It doesn't take much to create a deepfake video, all you need is a few pictures.

00:49 According to experts, deepfakes have soared tremendously in the last few years. So why

00:53 is this technology particularly bad news for women? Well, it's been weaponised to create

00:58 pornographic video content. In 2019, DeepTrace published an alarming report. This AI specialised

01:05 company estimated that porn content makes up 96% of deepfakes. And what's shocking is

01:11 that the vast majority of these videos were created without the consent of the featured

01:15 person. When it all started on internet forums in 2017, you needed serious skills to create

01:22 deepfake videos. The main targets were celebrity women due to the wide breadth of public

01:27 available pictures. As this technology becomes increasingly available today, any woman is

01:33 at risk of becoming a potential target. This creates a risk of serious psychological and

01:37 social repercussions. Well, it's horrifying and shocking to see yourself depicted in a

01:43 way that you didn't consent to. This is violating and this is dehumanising and the reality is

01:55 that this could impact a person's employability. This could impact a person's interpersonal

02:03 relationships and mental health. One of the most disturbing trends I see on forums of

02:10 people making this content is that they think it's a joke or they don't think it's serious

02:15 because the outputs aren't hyper realistic, not understanding that for victims, this is

02:20 still really, really painful and traumatic. If some creators make deepfakes as a joke,

02:28 others have malicious intentions. To tackle this, the FBI published a warning in June

02:33 this year regarding deepfakes being used for extortion purposes. Deepfakes have also been

02:39 used to discredit the work of women. For example, investigative journalist Rana Ayyub and activist

02:45 Kate Isaacs were both the victims of smear campaigns which weaponised deepfakes. One

02:51 of the big issues is that it's extremely difficult to control or even to detect deepfakes. We

02:56 want to tell people that they can do something to detect fakes or to fight back against this

03:01 problem. But the problem with that approach is that realistically, the individual and

03:08 the naked eye of the individual is just not going to be a reliable marker for spotting

03:15 fakes, you know, even now, but particularly moving into the future as the outputs get

03:19 better in quality. Around the world, only a few governments have taken up the issue

03:24 at a legislative level. So while the boom in artificial intelligence has been fun for

03:28 a lot of people, it represents serious risks for women online. But for now, there's still

03:34 a long way to go as lawmakers must catch up with technological innovations. There has

03:39 to be some sort of global response from governments, law enforcement, from people on the ground,

03:49 victim communities, so that there is accountability. Accountability for people where they can't

03:56 just ruin someone's life and get away with it and face no repercussions.

04:03 [BLANK_AUDIO]