Meta says that its current approach to labeling AI-generated content is too narrow and that it will soon apply a "Made with AI" badge to a broader range of videos, audio and images. Starting in May, it will append the label to media when it detects industry-standard AI image indicators or when users acknowledge that they’re uploading AI-generated content. The company may also apply the label to posts that fact-checkers flag, though it's likely to downrank content that's been identified as false or altered.

The company announced the measure in the wake of an Oversight Board decision regarding a video that was maliciously edited to depict President Joe Biden touching his granddaughter inappropriately. The Oversight Board agreed with Meta's decision not to take down the video from Facebook as it didn't violate the company's rules regarding manipulated media. However, the board suggested that Meta should “reconsider this policy quickly, given the number of elections in 2024.”

Meta says it agrees with the board's "recommendation that providing transparency and additional context is now the better way to address manipulated media and avoid the risk of unnecessarily restricting freedom of speech, so we’ll keep this content on our platforms so we can add labels and context." The company added that, in July, it will stop taking down content purely based on violations of its manipulated video policy. "This timeline gives people time to understand the self-disclosure process before we stop removing the smaller subset of manipulated media," Meta's vice president of content policy Monika Bickert wrote in a blog post.

Meta had been applying an “Imagined with AI” label to photorealistic images that users whip up using the Meta AI tool. The updated policy goes beyond the Oversight Board's labeling recommendations, Meta says. "If we determine that digitally-created or altered images, video or audio create a particularly high risk of materially deceiving the public on a matter of importance, we may add a more prominent label so people have more information and context," Bickert wrote.

While the company generally believes that transparency and allowing appropriately labeled AI-generated photos, images and audio to remain on its platforms is the best way forward, it will still delete material that breaks the rules. "We will remove content, regardless of whether it is created by AI or a person, if it violates our policies against voter interference, bullying and harassment, violence and incitement, or any other policy in our Community Standards," Bickert noted.

Over the years, Engadget has been the target of a common SEO scam, wherein someone claims ownership of an image and demands a link bank to a particular website. A lot of other websites would tell you the same thing, but now the scammers are making their fake DMCA takedown notices and threats of legal action look more legit with the help of easily accessible AI tools. 

According to a report by 404Media, the publisher of the website Tedium received a "copyright infringement notice" via email from a law firm called Commonwealth Legal last week. Like older, similar attempts at duping the recipient, the sender said they're reaching out "in relation to an image" connected to their client. In this case, the sender demanded the addition of a "visible and clickable link" to a website called "tech4gods" underneath the photo that was allegedly stolen. 

Since Tedium actually used a photo from a royalty-free provider, the publisher looked into the demand, found the law firm's website, and upon closer inspection, realized that the images of its lawyers were generated by AI. As 404Media notes, the images of the lawyers had vacant looks in the eyes that's commonly seen in photos created by AI tools. If you do a reverse image search on them, you'll get results from a website with the URL generated.photos, which uses artificial intelligence to make "unique, worry-free model photos... from scratch." The publisher also found that the law firm's listed address that's supposed to be on the fourth floor of a building points to a one-floor structure on Google Street View. The owner of tech4gods said he had nothing to do with the scam but admitted that he used to buy backlinks for his website. 

This is but one example of how bad actors can use AI tools to fool and scam people, and we have to be more vigilant as instances like this will just likely keep on growing. Reverse image search engines are your friend, but they may not be infallible and may not always help. Deepfakes, for instance, have become a big problem in recent years, as bad actors continue to use them to create convincing videos and audio not just to scam people, but also to spread misinformation online. 

Meta AI is consistently unable to generate accurate images for seemingly simple prompts like “Asian man and Caucasian friend,” or “Asian man and white wife,” The Verge reports. Instead, the company’s image generator seems to be biased toward creating images of people of the same race, even when explicitly prompted otherwise.

Engadget confirmed these results in our own testing of Meta’s web-based image generator. Prompts for “an Asian man with a white woman friend” or “an Asian man with a white wife” generated images of Asian couples. When asked for “a diverse group of people,” Meta AI generated a grid of nine white faces and one person of color. There were a couple occasions when it created a single result that reflected the prompt, but in most cases it failed to accurately depict the prompt.

As The Verge points out, there are other more “subtle” signs of bias in Meta AI, like a tendency to make Asian men appear older while Asian women appeared younger. The image generator also sometimes added “culturally specific attire” even when that wasn’t part of the prompt.

It’s not clear why Meta AI is struggling with these types of prompts, though it’s not the first generative AI platform to come under scrutiny for its depiction of race. Google’s Gemini image generator paused its ability to create images of people after it overcorrected for diversity with bizarre results in response prompts about historical figures. Google later explained that its internal safeguards failed to account for situations when diverse results were inappropriate.

Meta didn’t immediately respond to a request for comment. The company has previously described Meta AI as being in “beta” and thus prone to making mistakes. Meta AI has also struggled to accurately answer simple questions about current events and public figures.

The craze around Facebook Live might be a thing of the past, but Meta is still trying to make the platform video-friendly. The company has announced a new video player for uniformly displaying Reels, longer content and Live videos on the Facebook app. 

One of the biggest shifts is that all of Facebook's videos will now appear full-screen — even landscape-oriented ones. Videos will automatically play vertically, but you can now turn your phone on its side to watch most horizontal content across your entire device. 

Like many videos on TikTok, Facebook will now offer a slider at the bottom of the screen, letting you quickly move through the video. The update also brings some of the same features streamers like Netflix offer in their apps, such as the option to jump forward or backward by 10 seconds. Meta claims that you will now get "more relevant video recommendations" of all lengths appearing on the video tab and in your feed. The company will also be increasing the number of Reels shown on Facebook. 

The video player is rolling out now to Android and iOS users in the United States and Canada, with the new controls launching in the next few weeks. The entire update should be available globally in the coming months.

X has named a new head of safety nearly a year after the last executive in the position resigned. The company said Tuesday that it had promoted Kylie McRoberts to Head of Safety and hired Yale Cohen as Head of Brand Safety and Advertiser Solutions.

The two will have the unenviable task of leading X’s safety efforts, including its attempts to reassure advertisers that the platform doesn’t monetize hate speech or terrorist content. The company said earlier this year it planned to hire 100 new safety employees after previously cutting much of its safety staff.

Head of safety has been a particularly fraught position since Elon Musk took over the company previously known as Twitter. Musk has previously clashed with his safety leads and McRoberts is the third person to hold the title in less than two years. Previously, Yoel Roth resigned shortly after the disastrous rollout of Twitter Blue in 2022. Roth was replaced by Ella Irwin, who resigned last year after Musk publicly criticized employees for enforcing policies around misgendering.

Not much is known about McRoberts, but she is apparently an existing member of X’s safety team (her X account is currently private and a LinkedIn profile appears to have been recently deleted). “During her time at X, she has led initiatives to increase transparency in our moderation practices through labels, improve security with passkeys, as well as building out our new Safety Center of Excellence in Austin,” X said in a statement.

Spotify Premium users in Canada, Ireland and New Zealand will have access to 15 hours of monthly audiobook listening at no extra cost starting on April 9. Subscribers in the US, UK and Australia have had access to this perk for several months.

The Premium audiobook catalog now includes more than 250,000 titles. That's a notable increase from the 200,000 audiobooks that were in the library as of late 2023. So when you could use a change from the millions of songs and podcasts on Spotify, you'll have a ton of books to choose from.

Those who hit the 15-hour limit can add more audiobook listening time in 10-hour top ups. In the new markets, the extra listening time costs CAD $14.99, IRE €12.99 or NZD $19.99, per TechCrunch.

Since last month, Spotify has offered an audiobook-only subscription plan in the US. At $10, it's $1 per month less than Spotify Premium for the same 15 hours of audiobook listening time. Still, depending on the lengths of books that you listen to, this plan might prove better value than Audible, which grants you one audiobook credit per month for $15. That said, unused audiobook listening time on Spotify doesn't carry over to the next month.

On Monday, OpenAI began opening up ChatGPT to users without an account. It described the move as part of its mission to “make tools like ChatGPT broadly available so that people can experience the benefits of AI.” It also gives the company more training data (for those who don’t opt out) and perhaps nudges more users into creating accounts and subscribing for superior GPT-4 access instead of the older GPT-3.5 model free users get.

I tested the instant access, which — as advertised — allowed me to start a new GPT-3.5 thread without any login info. The chatbot’s standard “How can I help you today?” screen appears, with optional buttons to sign up or log in. Although I saw it today, OpenAI says it’s gradually rolling out access, so check back later if you don’t see the option yet.

OpenAI says it added extra safeguards for accountless users, including blocking prompts and image generations in more categories than logged-in users. When asked for more info on what new categories it’s blocking, an OpenAI spokesperson told me that, while developing the feature, it considered how logged-out GPT-3.5 users could potentially introduce new threats.

The spokesperson added that the teams in charge of detecting and stopping abuse of its AI models have been involved in creating the new feature and will adjust accordingly if unexpected threats emerge. Of course, it still blocks everything it does for signed-in users, as detailed in its moderation API.

You can opt out of data training for your prompts when not signed in. To do so, click on the little question mark to the right of the text box, then select Settings and turn off the toggle for “Improve the model for everyone.”

OpenAI says more than 100 million people across 185 countries use ChatGPT weekly. Those are staggering numbers for an 18-month-old service from a company many people still hadn’t heard of two years ago. Today’s move gives those hesitant to create an account an incentive to take the world-changing chatbot for a spin, boosting those numbers even more.

LinkedIn is testing a new feed of TikTok-like vertical videos. The feature hasn’t been publicly announced but it’s been spotted by users in recent days and the company confirmed the tests to TechCrunch.

According to a screenshot shared by Instagram employee Jenny Eishingdrelo and a video posted to LinkedIn by influencer marketing exec Austin Null, the new feed will appear in a separate “video” tab in the LinkedIn app. Users will be able to scroll vertically to move between clips, much like TikTok or Instagram Reels.

It’s not the first time the company has hopped on a trendy format. LinkedIn previously experimented with a Stories feature for disappearing posts. That feature lasted less than a year, though the professional network hinted at the time that it wasn’t done with its video experiments, saying it was working “to evolve the Stories format into a reimagined video experience across LinkedIn.”

Presumably, LinkedIn is hoping the feed will showcase content from its ranks of professional creators and thought leaders, many of whom are already posting video to their feeds. However, it’s not clear how many of the site’s users are interested in a dedicated video feed for workplace-related content.

Instagram is working on a feature that would recommend Reels to you and a friend based on videos you've shared with each other and your individual interests. Reverse engineer Alessandro Paluzzi unearthed the feature, which is called Blend. Instagram confirmed to TechCrunch that it's testing Blend internally and it hasn't started trialing it publicly. It may be the case that Blend never sees the light of day, though it's always intriguing to find out about the ideas Instagram is toying with.

The platform hasn't revealed more details about how Blend will work, though the idea seems to be that Instagram users and one of their besties will discover new Reels together instead of one of them finding a video they like and DMing it to the other. It would make sense for Blend to have an indicator that the other person has already seen a particular Reel so that the two people who have access to the feed can start chatting about it. 

TikTok doesn't have a feature along these lines, as TechCrunch notes, so Blend could give Instagram an advantage when it comes to folks who like to check out short-form videos together. As with many of the other features platforms of this ilk introduce, Blend fundamentally seems to be about increasing engagement.

YouTube’s TikTok competitor, Shorts, is becoming a more significant part of the company’s monetization program. The company announced that more than a quarter of channels in its Partner Program are now earning money from the short-form videos.

The milestone comes a little more than a year after YouTube began sharing ad revenue with creators making Shorts. YouTube says it currently has more than 3 million creators around the world in the Partner Program, which would imply the number of Shorts creators making money from the platform is somewhere in the hundreds of thousands.

Because ads on Shorts appear between clips in a feed, revenue sharing for Shorts is structured differently than for longer-form content on YouTube. Ad revenue is pooled and divided among eligible creators based on factors like views and music licensing. The company has said this arrangement is far more lucrative for individuals than traditional creator funds.

So far though, it’s unclear just how much creators are making from Shorts compared with the platform’s other monetization programs. YouTube declined to share details but said the company has paid out $70 billion to creators over the last three years.

Shorts’ momentum could grow even more in the coming months. TikTok, which itself has been trying to compete more directly with YouTube by encouraging longer videos, is facing a nonzero chance that its app could be banned in the United States. Though that outcome is far from certain, YouTube would almost certainly attract former TikTok users and creators.