AI tools may help proliferate deepfake porn. What's being done to … – USA TODAY

NEW YORK — Artificial intelligence imaging can be used to create art, try on clothes in virtual fitting rooms or help design advertising campaigns. 
But experts fear the darker side of the easily accessible tools could worsen something that primarily harms women: nonconsensual deepfake pornography.
Deepfakes are videos and images that have been digitally created or altered with artificial intelligence or machine learning. Porn created using the technology, often targeting online influencers, journalists and others with a public profile exists across a plethora of websites. Some offer users the opportunity to create their own images – essentially allowing anyone to turn whoever they wish into sexual fantasies without their consent, or use the technology to harm former partners.
Experts say the problem could get worse with the development of generative AI tools that are trained on billions of images from the internet and spit out novel content using existing data.
Here’s how AI models and online platforms are trying to curb that.
She discovered a naked video of herself, but it wasn’t her:The trauma of deepfake porn
AI-generated images already fool people:Why experts say they’ll only get harder to detect
Governing the internet is next to impossible when countries have their own laws for content that’s sometimes made halfway around the world, but some AI models say they’re already curbing access to explicit images. 
OpenAI says it removed explicit content from data used to train the image-generating tool DALL-E, which limits the ability of users to create those types of images. The company also filters requests and says it blocks users from creating AI images of celebrities and prominent politicians. Midjourney, another model, blocks the use of certain keywords and encourages users to flag problematic images to moderators.
Meanwhile, the startup Stability AI rolled out an update in November that removes the ability to create explicit images using its image generator Stable Diffusion. Those changes came following reports that some users were creating celebrity-inspired nude pictures using the technology.
Stability AI spokesperson Motez Bishara said the filter uses a combination of keywords and other techniques like image recognition to detect nudity and returns a blurred image. But it’s possible for users to manipulate the software and generate what they want since the company releases its code to the public. Bishara said Stability AI’s license “extends to third-party applications built on Stable Diffusion” and strictly prohibits “any misuse for illegal or immoral purposes.”
Some social media companies have also been tightening up their rules to better protect their platforms against harmful materials.
TikTok said last month all deepfakes or manipulated content that show realistic scenes must be labeled to indicate they’re fake or altered in some way, and that deepfakes of private figures and young people are no longer allowed. Previously, the company had barred sexually explicit content and deepfakes that mislead viewers about real-world events and cause harm.
The gaming platform Twitch also recently updated its policies around explicit deepfake images after a popular streamer named Atrioc was discovered to have a deepfake porn website open on his browser during a livestream in late January. The site featured phony images of fellow Twitch streamers.
Twitch already prohibited explicit deepfakes, but now showing a glimpse of such content – even if it’s intended to express outrage –“will be removed and will result in an enforcement,” the company wrote in a blog post. And intentionally promoting, creating or sharing the material is grounds for an instant ban.
She paid a photographer $100 for photos:Now she’s the face of an erotic novel, breast reduction ads
Being harassed online? Prevent cyberstalking with these tips
Other companies have also tried to ban deepfakes from their platforms, but keeping them off requires diligence.
Apple and Google said recently they removed an app from their app stores that was running sexually suggestive deepfake videos of actresses to market the product. Research into deepfake porn is not prevalent, but one report released in 2019 by the AI firm DeepTrace Labs found it was almost entirely weaponized against women and the most targeted individuals were western actresses, followed by South Korean K-pop singers.
The same app removed by Google and Apple had run ads on Meta’s platform, which includes Facebook, Instagram and Messenger. Meta spokesperson Dani Lever said in a statement the company’s policy restricts both AI-generated and non-AI adult content and it has restricted the app’s page from advertising on its platforms.
In February, Meta, as well as adult sites like OnlyFans and Pornhub, began participating in an online tool, called Take It Down, that allows teens to report explicit images and videos of themselves from the internet. The reporting site works for regular images, and AI-generated content – which has become a growing concern for child safety groups.
How do I stop Apple from showing my location? Halt Google, others, from seeing your data
Cybersecurity tech tips:5 secrets only pros and hackers know

source

Leave a Comment