Gaurav Misra spent about $50,000 to build and run a television news operation for a day, without hiring a single journalist.

About $30,000 of that went to Claude Code credits as Misra, the CEO of AI video company Mirage, built the operation himself. The result was the Mirage News Network, a 24-hour experiment in using artificial intelligence to produce something resembling a traditional television news channel.

The broadcast drew about 50,000 viewers on X and the livestream post generated 824,000 impressions, according to Variety. Mirage paid several accounts to promote the stream, meaning some of that reach was purchased.

The Mirage News Network ran for 24 hours on Aug. 11. Four AI-generated anchors, styled after television broadcasters from the 1980s and given names including “Tom Callahan” and “Marcus Sterling,” delivered stories licensed from Reuters that were already several days old.

Mirage used its Avatar X model to create the anchors, while ChatGPT’s image generator and Google’s Gemini text-to-speech technology helped produce other parts of the broadcast. The humans who appeared on screen came from Reuters video reports and stock footage.

Even the guests were partly synthetic.

Mirage recruited more than a dozen people, many of them venture capital professionals, who allowed the company to use their likenesses and approved scripts for their AI doubles to deliver. Billy McFarland, the organizer of the failed Fyre Festival, was among them.

Mirage does not employ journalists. Misra, a former Snap and Microsoft engineer, told Variety that he instructed the system to follow “general journalistic standards,” rather than a specific set of guidelines such as AP style or the Society of Professional Journalists’ code of ethics.

The company did disclose its use of AI. An on-screen graphic identified the anchors, graphics, guests and advertisements as artificially generated.

But the experiment also showed some of the problems that can emerge when a news operation has little human oversight. At one point, the broadcast incorrectly identified a woman appearing in a Reuters report as President Donald Trump. Mirage later posted a correction on X rather than correcting the error during the broadcast.

Dan Axelrod, chair of the Society of Professional Journalists’ ethics committee, told Variety he considered Mirage’s experiment a news delivery project rather than a news network. He called it “a false and misguided devaluing and deskilling of journalism.”

His committee proposed revisions to SPJ’s code of ethics the same week, as the organization considers how longstanding journalistic principles should apply as newsrooms adopt new technology.

Even some people inside Mirage treated the project more like an experiment than the future of television news. Engineer Jason Silberman described it on X as what happens “when the CEO starts tokenmaxxing too hard and spends $50k in a day on a marketing stunt.”

There is also little evidence so far that large numbers of Americans want AI to become their primary source of news.

A Gallup survey released in May found that 7% of Americans said they relied on AI for news to at least a moderate degree. Nearly 4 in 10 said a news organization’s use of AI would make them trust its reporting less.

For publishers, Mirage may be more useful as an experiment in the economics of automated news than as a model for replacing cable television.

Smaller AI-native news operations are already testing versions of the idea. Runtime Wire operates with one human editor and says its system rejects 97.7% of the stories it considers, leaving at least some editorial judgment with a person.

Disclosure is another unresolved issue. Mirage prominently labeled its synthetic broadcast, but research into AI labeling has shown that disclosure alone does not guarantee audiences will notice or understand those labels.

At the same time, the financial pressure pushing publishers toward automation is real. Bauer Media, for example, stopped commissioning freelance fiction writers for Take a Break as it began using AI to help produce stories internally.

Misra told Variety that AI-generated newscasts are “probably inevitable in the future.”

The post Mirage’s $50,000 AI news channel runs 24/7 without journalists appeared first on The Media Copilot.

Tap the headphones icon at the top of a New York Times or New Yorker story and you may hear it read by an automated voice. The feature has spread across major news organizations, as publishers use AI-generated narration to give readers another way to consume their journalism.

In a Columbia Journalism Review analysis, contributing writer Jem Bartholomew examines how The New York Times, The Washington Post, The Wall Street Journal, The New Yorker, The Atlantic and Vanity Fair have adopted automated text-to-speech tools. ElevenLabs, the AI voice company that partnered with The New Yorker in 2023, said this year that it had surpassed $500 million in annual recurring revenue. The company offers more than 10,000 artificial voices and has held early talks with investors about a secondary offering that would value it at $22 billion.

The technology is already drawing significant use. About 20% of New Yorker subscribers listen to narrated stories, deputy editorial director Monica Racic told CJR. At The Wall Street Journal, the “Read to Me” feature, developed through a partnership with Microsoft that began in 2020, was used about 5 million times over the past year and had a 65% completion rate.

New Yorker editor David Remnick told Semafor that an ElevenLabs voice sounded “pretty damn good” and was “nearly instantaneous” to produce. ElevenLabs CEO Mati Staniszewski told Al Jazeera in February that the company’s technology could produce speech based on the contextual understanding of written text.

But CJR’s Bartholomew found limits in the technology. AI narrators can rush through passages that call for pauses, emphasize unexpected words and use simulated breaths or rising intonation at unnatural moments. The result can sound convincingly human while still feeling artificial.

Bartholomew found the difference most striking when he compared an AI narration of a 1997 New Yorker profile of Donald Trump by Mark Singer with Singer’s own reading of the piece for the audiobook version of his 2016 book, Trump and Me. Singer died in June.

The synthetic voice maintained a steady pace through passages that depended on irony, sadness and changes in tone. Singer’s original reading conveyed those shifts; the AI version largely did not. For Bartholomew, the comparison showed what can be lost when a piece of writing shaped by a human writer is delivered by a synthetic voice.

Despite the loss of tone and nuance, accessibility is one of the clearest arguments for AI narration. A Times spokesperson told CJR that the newspaper has experimented with automated voices for years as a way to make its journalism more accessible, including for people who are blind or have low vision. ElevenLabs says its impact program provides free licenses to nonprofits and people with accessibility needs.

Publishers also see narration as an additional way to reach audiences rather than a replacement for written journalism. Taneth Evans, the Journal’s head of digital, told CJR that the feature is “additive” and gives readers another way to engage with the newspaper’s work.

The New Yorker has drawn a line around fiction. The magazine does not use AI to narrate its fiction, which is read by authors whenever possible because the reading is treated as a performance.

AI narration gives publishers another way to put journalism in front of audiences wherever they are. But as synthetic voices become more common, newsrooms face a more consequential question: How much of a story’s meaning survives when the human voice is taken out of the telling?

The post The rise of the AI newsreader: ElevenLabs eyes $22 billion valuation appeared first on The Media Copilot.

YouTube trust and safety chief Matt Halprin sat down for a Creator Insider video last week to make more explicit what the platform means by “inauthentic content.” The result, rolled out July 16, is a set of clarifications that spell out three specific buckets of video that can no longer earn money through the YouTube Partner Program.

The update isn’t a brand-new rule. As TechCrunch reported, YouTube already moved last year to stop creators from making revenue off mass-produced, repetitive videos that AI tools make cheap and fast to churn out. This latest change adds detail to those existing guidelines rather than replacing them.

The first category is generic, repetitive or template-based content. Halprin described channels stuffed with cookie-cutter clips made through AI, CGI or templates that barely change from one video to the next.

The second targets what YouTube calls off-putting content, meaning videos built to distress or emotionally manipulate viewers into clicking. Halprin’s example: an animal shown in distress before someone conveniently arrives to rescue it.

“We’ve heard from our viewers that that’s not something that they like,” he said. Channels dedicated to this content lose Partner Program access whether or not AI made the videos.

The third bucket goes after AI personas, which are AI-made representations of real people.

Halprin was careful not to frame AI as the villain. “AI can actually allow people to make a lot of videos,” he said. “Sometimes those videos are great, and it really enhances creativity.” The same tools, he added, also let people spit out large volumes of near-identical clips with no narrative arc, which is the content farming YouTube wants out of its monetization program.

The Partner Program, which pays creators through ads and subscriptions, is central to YouTube’s business, and the platform now pulls in more ad revenue than Disney, Paramount and Warner Bros. Discovery. Letting the feed fill with low-quality AI output risks the viewer trust that keeps those ad dollars flowing.

For publishers and newsrooms experimenting with AI-assisted video, the takeaway is about intent, not tools. YouTube isn’t penalizing AI use itself. It’s penalizing volume without originality, manipulation for clicks and synthetic voices on topics where accuracy carries real stakes.

The biggest unanswered question is where YouTube will draw the line. Halprin said channels with too much repetitive, low-effort or manipulative content will lose monetization, but he did not define the threshold. The clarified policy applies immediately to all YouTube Partner Program members.

The post YouTube tightens monetization rules around AI slop appeared first on The Media Copilot.

OpenAI’s first piece of consumer hardware is reportedly a screenless, mobile AI companion that syncs with ChatGPT and is being pitched internally as a humanlike AI companion that lives in the home.

That description comes from a Bloomberg report published Tuesday, which TechCrunch summarized the same day. The device is still under development, and its sources describe it as something well outside the standard smart-speaker mold.

The speaker reportedly has a personality and can learn about its owner over time to deliver more personalized responses. It would tap into a user’s digital life, pulling from sources like email. And it apparently includes mechanical elements that can move on their own, designed to feel like a companion and act as a physical manifestation of ChatGPT.

OpenAI announced prototypes for hardware in November 2025, with earlier rumors pointing to a phone that would compete directly with Apple. The company also acquired io, the hardware startup co-founded by former Apple design chief Jony Ive, in a deal valued at roughly $6.5 billion. Bloomberg reports the speaker was built with help from ex-Apple engineers who worked on the iPhone and Mac.

That pedigree is now a legal liability. Apple sued OpenAI last week, accusing it of stealing trade secrets and calling the allegations “the tip of the iceberg.” OpenAI denies wrongdoing. Its sources told Bloomberg the new product veers significantly from anything Apple sells today.

OpenAI is not alone in chasing this category. Hark, the AI lab founded by Brett Adcock, raised an oversubscribed $700 million Series A in May at a $6 billion valuation to build proprietary models paired with custom hardware it calls a universal interface between humans and machines. Hark has yet to disclose the device’s design, highlighting the surge of investment flowing into AI hardware long before products ship.

For newsrooms and publishers, a screen-free device that pulls from email and learns a user’s habits raises the same questions as any voice-first platform. If people increasingly ask a companion device for news instead of opening an app or a browser, distribution shifts again, and this time to a surface with no visible headlines, no scannable feed and no obvious place for a byline.

The referral traffic problem publishers already face with chatbots gets sharper when the interface is spoken. The reported device would extend a broader shift in news discovery, making AI assistants an even more prominent intermediary between publishers and their audiences.

OpenAI has not announced the product, its price or when it will ship. Still, the reported device points to the company’s broader strategy of extending ChatGPT beyond screens and into the home.

The post OpenAI’s first hardware device is reportedly a screenless speaker that moves appeared first on The Media Copilot.

YouTube is moving from voluntary disclosure to automatic detection when it comes to AI-generated content.

The platform announced Wednesday that it will begin automatically applying AI-generated content labels to videos that feature “significant photorealistic AI use”—even if the creator never disclosed it. As noted by Variety, the change marks a notable escalation of YouTube’s approach to AI transparency, which previously relied entirely on creators self-reporting their use of generative AI tools.

Previously, YouTube labeled AI-generated content only when creators voluntarily disclosed it in their video settings. Now the company is rolling out an internal detection system that will flag videos even without creator admission. Creators can dispute incorrect labels through YouTube Studio, but YouTube says the labels will “remain permanent” in certain cases, including content created using YouTube’s own AI tools like Veo or Dream Screen, and any video carrying C2PA metadata (standards from the Coalition for Content Provenance and Authenticity) indicating full AI generation.

The labels are also getting a more prominent placement. Previously buried in expanded descriptions, AI labels on long-form videos will now appear directly below the video player, above the description. For YouTube Shorts, the label will appear as an overlay directly on the video itself.

“The goal here is context at a glance,” said Rene Ritchie, YouTube head of editorial and creator liaison, in a video explaining the changes. “If it looks real but was made with AI, viewers will know immediately.” Ritchie emphasized that the labels do not affect monetization or recommendation algorithms — “This is purely about giving viewers the right information at the right time.”

This push for better AI disclosure follows a broader problem: why AI content labels keep failing the people who need them most. Content Credentials, the metadata-based standard designed to track an image’s AI origins, has existed for years, but social platforms have been inconsistent about adopting it, often stripping out the very metadata that makes the system work. YouTube’s move toward automatic detection is an attempt to close that gap, even if the underlying standards remain patchily implemented.

The move comes alongside YouTube’s expanded likeness-detection program, now available to all creators 18 and older, which helps users identify and request removal of AI-altered facial likeness content.

The post YouTube Will Auto-Label AI-Generated Videos, and Make Those Labels Harder to Miss appeared first on The Media Copilot.

Google has removed dozens of AI-generated videos featuring Disney-owned characters from YouTube after receiving a cease and desist letter from the studio on Wednesday.

The videos included Mickey Mouse, Deadpool, and characters from Star Wars and The Simpsons. Many were created using Veo, Google’s own AI video generation tool, according to Gene Maddaus at Variety.

The timing is notable. Disney sent the letter shortly before announcing a deal to license 200 characters to OpenAI, allowing Sora users to create short AI clips with official permission. The message is clear: unlicensed use will be fought, but licensing deals are on the table.

Disney’s letter included a lengthy list of protected characters from Frozen, Moana, Toy Story, Iron Man, Lilo & Stitch, and Winnie the Pooh. The studio also demanded Google implement safeguards to prevent its AI tools from generating Disney characters and stop using Disney content to train its models.

Google responded carefully. “We have a longstanding and mutually beneficial relationship with Disney, and will continue to engage with them,” the company said in a statement. Google pointed to its existing copyright tools, including Content ID for YouTube.

Why this matters for newsrooms: This is the clearest example yet of the carrot-and-stick approach that major content owners are taking with AI companies. As publishers and media companies evaluate their own AI licensing strategies, Disney’s model offers a template: aggressive enforcement against unauthorized use combined with willingness to strike commercial deals.

The removed videos now redirect to a message: “This video is no longer available due to a copyright claim by Disney.”

The post Disney finds Google’s lack of licensing disturbing, forces takedown of AI videos appeared first on The Media Copilot.