Mirage Puts an AI News Network on the Air, Streaming Nonstop on X

# Mirage Puts an AI News Network on the Air, Streaming Nonstop on X

On the morning of Aug. 12, viewers who clicked into a livestream on X found something that looked, at a glance, like any cable news channel: crisp anchors at a desk, lower-third headlines, packages cut with wire footage. What they would not find, anywhere in the broadcast, was a human being on camera. The anchors were synthetic. The scripts, the voices, the pacing between segments — all of it was generated by software. With that, Mirage, the AI video company formerly known as Captions, declared it had launched the world's first AI news network, streaming nonstop through its @TryMirage account.

The pitch is audacious in its simplicity. Rather than route viewers to a new app or a standalone site, Mirage planted its channel directly on X, where an audience is already scrolling. The debut broadcast reached roughly 50,000 viewers, and the livestream post drew about 824,000 impressions, according to figures reported around the launch — modest by the standards of legacy television, but notable for a channel with no newsroom, no camera crew, and no on-air talent drawing a salary.

How the channel is built

The network is less a newsroom than a pipeline. According to Variety, the broadcast's four anchors were assembled by an AI agent stitching together a mix of tools: Mirage's own Avatar X model, which the news network was arguably launched to promote, along with ChatGPT's image generator and Google's Gemini text-to-speech feature. The result is a rotating cast of presenters who do not exist, reading news they did not report.

The news itself comes from elsewhere. The stories aired in the debut stream were days-old items licensed from Reuters' wire service. In a telling wrinkle, the only real people who appeared anywhere in the broadcast were the human subjects Reuters journalists had originally filmed and the stock footage packaged with those reports. The synthetic layer sits on top of conventional journalism — a machine-generated face and voice reciting work that human reporters did first.

That arrangement flows from what Mirage already builds. Founded in 2021 by former Snap engineers Gaurav Misra and Dwight Churchill, the company made its name on tools that automate short-form video creation, and its Mirage Studio platform generates AI actors and marketing spots for brands. The company raised $75 million in March 2026, bringing its total funding to roughly $175 million. The news network extends that same synthetic-media machinery from advertising into live journalism.

Who is this for?

Misra frames the effort as expansion rather than replacement. He has argued that there may be an audience of "everyday people" who would prefer a generated presenter to a living one for "a different angle" on the news, and that the technology "just lowers the barrier for more people to be able to participate in creating content." The vision is a world where standing up a news channel costs a software subscription rather than a broadcast license and a payroll.

The skeptics are not hard to find, and the numbers give them ammunition. A Gallup survey in May 2026 found that only 7% of Americans rely on AI to get news either "a great deal" or "a fair amount," while nearly 40% said a news organization's use of AI would reduce their trust in the information. Pew Research Center has found that nearly 60% of adults between 18 and 29, and 51% of those 30 to 49, believe at least some news produced by chatbots is inaccurate. Meanwhile the 2026 Reuters Institute Digital News Report described an "unsettled time," with falling news interest and trust even as social platforms, video creators, and AI chatbots absorb more of the audience. Launching an AI anchor into that climate is, at minimum, a bet against the prevailing mood.

The trust problem underneath

The core tension is not whether the technology works — clearly it produces watchable video — but whether audiences should trust a news product with no human accountable on screen. For now, Mirage sidesteps the sharpest risk by licensing established wire copy rather than having its models generate original reporting. That keeps the system away from AI's most notorious failure mode: hallucination, the tendency of generative models to state fabricated facts with total confidence. But it also exposes the thinness of the value proposition. If the journalism is Reuters' and the only new layer is a synthetic face, the network's real product is presentation, not reporting.

The deeper question is what happens if this model scales beyond licensed wire feeds. A pipeline that can turn text into a convincing anchor reading it is precisely the tool that makes synthetic misinformation cheap and fast. Mirage's channel, tethered to Reuters and clearly labeled, is the benign version. The same capability in less scrupulous hands is the reason media researchers have spent years building tools to detect exactly this kind of content.

What to watch

Three signals will tell whether Mirage has a business or a stunt. First, engagement over time: a launch-day audience of 50,000 is a curiosity number, not a franchise, and whether viewers return is the real test. Second, the sourcing: if Mirage moves from licensed wire packages toward AI-generated or AI-summarized original content, the trust and accuracy stakes rise sharply. Third, the platform relationship — Mirage's decision to build natively on X ties its fortunes to Elon Musk's platform and its shifting content rules. For an industry already fighting to hold public trust, an anchor that cannot be held accountable is a provocative place to plant a flag. Whether audiences accept a face that was never real may say as much about the state of trust in news as it does about the state of AI.