On Wednesday police commissioner Krissy Barratt confirmed some of the women would be arrested and charged. The potential charges included terrorism offences such as entering, or remaining in, declared areas, and crimes against humanity offences, such as engaging in slave trading.
It is far too soon to mention retirement, but astute observers of the space industry have noticed SpaceX’s workhorse Falcon 9 rocket is not launching as often as it used to.
The decline is modest so far, and it does not signal any problem at SpaceX or with the Falcon 9. Rather, it is a manifestation of SpaceX’s eagerness to shift focus to the much larger Starship rocket, an enabler of what the company wants to do in space: missions to land on the Moon and Mars, orbital data centers, and next-gen Starlink.
Elon Musk’s SpaceX conducted 165 launches with the Falcon 9 rocket (no Falcon Heavy missions) last year, up from 134 Falcon 9 and Falcon Heavy launches in 2024 and 96 Falcon flights in 2023. The company plans “maybe 140, 145-ish” Falcon launches in 2026, SpaceX president Gwynne Shotwell told Time earlier this year. “This year we’ll still launch a lot, but not as much,” she said. “And then we’ll tail off our launches as Starship is coming online.”
We’re beginning to see what the long, slow tail-off will look like. The changes are most apparent at Cape Canaveral, Florida, where SpaceX has launched the lion’s share of its rockets. Until last December, SpaceX launched Falcon 9s with regularity from two pads on Florida’s Space Coast—one at NASA’s Kennedy Space Center and another a few miles to the south on military property at Cape Canaveral Space Force Station.
SpaceX is transitioning the site at Kennedy, known as Launch Complex-39A, to launch Starships. LC-39A is out of the rotation for Falcon 9 launches, although it remains available for occasional flights of the more powerful triple-core Falcon Heavy. SpaceX launched the first Falcon Heavy in a year and a half last week from LC-39A, and a handful more Falcon Heavy flights are on tap later this year.
Activity at SpaceX’s oldest launch site, Space Launch Complex-40 at Cape Canaveral, is also waning. Last month, SpaceX retired one of its two Florida-based seagoing landing platforms from service for future use as a transporter to ferry Starships and Super Heavy boosters from SpaceX’s factory in South Texas to Florida. SpaceX is constructing a second Starship factory at Kennedy Space Center, but officials want to begin Starship flights from Florida before the factory is operational.
“Clippers” cut up podcasts, videos, and events into infinite shorter versions. How long can they ride the algorithms?
Earlier this year, after a tumultuous period serving as the former second-in-command at the FBI, Dan Bongino went back to what he is perhaps known best for: video podcasting. After Bongino exited the role in January, he began promotion for the return of his podcast, The Dan Bongino Show. He bought out a billboard in Times Square in New York; he dropped teaser videos for his first new episode in months. Bongino also deployed a more experimental promotional tactic, aimed at getting portions of his show in front of a wider audience. For this, he used clippers.
Clippers are largely anonymous social media accounts whose sole purpose is to rack up views. The accounts take a piece of longform content — an hours-long livestream, for example, or a podcast — and pull out the most exciting, controversial, or shocking moments. Sometimes the accounts are dedicated to clipping, but companies will also recruit accounts with existing followers. Clippers can be based anywhere in the world (one tech founder who uses clippers has described some of them as “hungry Slovakian teenagers”) while targeting English-speaking audiences.
After clippers get the source material that a brand wants to promote, they cut it down and blast their version into the open web. Hundreds or even thousands of clipping accounts might be sharing similar videos, all in competition with one another. You have perhaps learned about a TV show moment, a celebrity podcast appearance, or a new band via clippers without even realizing it; it just looks like someone sharing something. Clippers do not need to be affiliated in any real way with the subjects they are clipping, and the clipped content does not need to be creative, transformative, or even interesting. It is the cartilage of the internet, the placeholders for the algorithm to suck in and spit out.
According to a campaign listing on the service Clipping.net, The Dan Bongino Show started a 31-day campaign beginning the day after his podcast returned in February. There were few requirements or guidelines, only that clippers should pull moments from his new podcast episode, and include #danbongino in the video caption. The campaign ran across TikTok, Instagram, and YouTube, and clippers would be paid $150 for every 100,000 views (with funds dispersed via PayPal). A Discord message about The Dan Bongino Show pegged the budget at $2,000. Bongino’s team did not respond to a request for comment.
“It’s just a necessary marketing play that if you’re not doing you’re behind,” Clipping founder Anthony Fujiwara told The Verge in a message. “Clipping lets you abuse the algorithms of other platforms to grow your product exponentially.” Fujiwara says 62,000 clippers use his platform, earning $3,000 a month on average. Most are based in the US.
“We verify using their audiences as a metric for who we want to be a clipper,” he says. “Indian views don’t help anyone.”
If you are someone who wants attention, social media is just just another form of gambling in the age of algorithmic recommendation feeds. Creators and influencers can optimize their content or tweak titles and thumbnails, but ultimately they are all just pulling a virtual slot machine arm, hoping it will dispense views, engagement, and resultant revenue. For well over a decade, content creators have worked to reverse engineer “the algorithm.” Deploying clippers allows companies to gamble on content at scale, without paying a network of contractors upfront: Why bet once, when you could bet 50 times? Clipping is nothing new, despite the recent discourse around who uses it and why, and whether paying random accounts to share content promoting something is deceptive or manufacturing fake fandom. The reality is that more and more, the social internet is filled with clips, paid and unpaid, that stand in for the full-length podcast, video, film, album, or piece of writing. As online content increasingly becomes abstracted from the original work, what purpose does making the full version even serve?
It’s not just podcasters who hire what is essentially a personal army of microtask workers. Clipping.net also lists campaigns for TV shows like RuPaul’s Drag Race ($175 per 100,000 views) and Michael Carbonara, a candidate running for congress in Florida. (The instructions for the campaign note dictate “Your clips must NOT have Michael saying anything Anti-Trump / Anti-White House,” and note that AI-generated clips are acceptable, though.) The brief doesn’t include any instructions for disclosing it is paid content; the Federal Election Commission requires that digital content include disclaimers. Carbonara’s campaign did not respond to requests for comment. World of Wonder, the production company for RuPaul’s Drag Race, declined to comment.
On Vyro, another clipping service launched by MrBeast, Perplexity launched a campaign in early April centered around Joe Rogan’s use of AI. (Perplexity is a sponsor of Rogan’s show.) Clippers were instructed to make content based on Rogan discussing AI with guests like Bradley Cooper and Johnny Knoxville, with the AI company specifically mentioned. The campaign ran across Instagram and TikTok, paying $1.20 per one thousand views on a video — and came with more requirements. Accounts were required to have more than 10,000 followers to submit clips, and all posts were to include #PoweredByPerplexity and #sponsored (many clipping campaigns have no disclosures that the content is paid). Reached via email, Perplexity distanced itself from the clipping company, with spokesperson Jesse Dwyer saying Perplexity “has no knowledge” of Vyro and “takes any unauthorized use of the Perplexity name or logo very seriously.” When asked to confirm Perplexity had not run or authorized clipping campaigns, Dwyer stopped responding to The Verge. Vyro directed me to Evangelist, a platform that connects brands with clip farms, but the company declined to comment. If clips really are the standard for marketing — a tool that everyone uses, that is at this point old news — why the secrecy?
One of the biggest beneficiaries of the clip economy is Clavicular (real name: Braden Peters), a 20-year-old streamer who has enjoyed an unprecedented come-up thanks to short videos of him going viral. He has received mainstream news coverage of his maniacal focus on his appearance, in a fringe subculture known as “looksmaxxing.” He has thrown around racist slurs; hit his face with a hammer, saying he was literally chiseling his bones; and sang along to Ye’s song “Heil Hitler” with other right wing and manosphere influencers. You probably have never watched one of Peters’ hours-long streams — but I bet you’ve seen clips of them.
According to figures posted by Peters, more than 1,600 clippers farmed out content of him between March and April, posting nearly 70,000 videos that accumulated more than 2 billion views. They clipped him on a fake date with another influencer, in nightclubs with a rotating cast of women, and apparently being consensually choked until he convulsed.
“Everyone hating but as predicted the clip went giga viral,” Peters wrote on X of the choking stunt.
The clip ecosystem has a way of elevating even unknown personalities, reaching millions of people who will potentially never see where the original material clips come from. Peters, for example, has around 337,000 followers on Kick, a fraction of the following others command on the platform. Relatively few people watch his streams live, which are, frankly, uneventful — except for the moments he’s manufactured to be clipped later.
You do not even need to hire people to clip your content: many will do it for free, or you can just recycle everything you produce as clips. TBPN, the three-hour-long podcast popular among a subset of the Silicon Valley tech industry, gets only a few thousand views on YouTube, but most people watch the show via disembodied segments — clips of each guest — on X. OpenAI recently acquired the show, which generated millions of dollars via flashy, maximalist ads on the livestreams. As part of the OpenAI deal, TBPN will “wind down” the advertising, according to The Wall Street Journal.
Clippers by and large add nothing of substance to the original work — no analysis, no response, not even filters and music and cuts that fan edits often have. It is the most boring kind of content, spliced purely for the algorithm. And when it comes to paid clippers, it is hard to separate the aesthetics from the often very young editors who churn out content. Often, the only edit made to videos is a solid border around a clip with a clickbait-y few lines of text: “Joe Rogan talking about AI sounding too real 😨” one video on Instagram reads. The video is a clip from Rogan’s interview with Knoxville and includes the tag #PoweredByPerplexity and #sponsored, as per the clipping campaign rules. Another Instagram account that last posted in 2017 when it was sharing calligraphy videos appears now to be part of a clip farm, sharing Call of Duty clips tagged as sponsored with captions like “This might be the smoothest sniping has ever been in a battle royale 👀.” The original clip that made Clavicular into a viral personality was nothing special, simply a short video with the Kick logo in it. The caption, filled with words that to the average person meant nothing, was the actual growth hack: “Clavicular ran into a frat leader at ASU and got brutally frame mogged by him👀😂”
Now that the existence of clipping has hit mainstream consciousness, the strategy is being touted as the future of building platforms and growing a business online. Clipping is undeniably effective at generating views, which in the era of shortform video largely means the number of times a post comes up on a user’s feed (as opposed to requiring a certain watch time, for example). But whether clipping actually builds a meaningful and resilient audience is, to me, still unproven. The Rogan and Knoxville clip on Instagram, for example, generated 272,000 views, but almost no engagement: just over 700 likes, 14 comments, and 10 reposts. Did Perplexity’s clipping campaign meaningfully affect its business, or Rogan’s viewership? Did it change anyone’s mind about AI? Or was it simply the connective tissue between one scroll and the next, a piece of media that popped up and was forgotten just as quickly?
TikTok’s ascendance during the pandemic made shortform video everyone’s problem (or solution, depending on how you look at it). In less than a decade’s time, we are in a second pivot to video phase that now stretches beyond news organizations. Podcasts have turned into video talk shows hosted by journalists, influencers, comedians, and nuns. Political strategy firms flood Instagram, TikTok, and X with clips of politicians at events saying something shocking or stupid, knowing the clip itself is the news. Even I, a features writer, participate in the clip-ification of our digital lives: I make shortform videos explaining my stories to TikTok audiences who will mostly not end up reading the original piece. Some of it is out of necessity — if you make something for public consumption, the devil’s bargain is you have to promote it. But overindexing on the clipped version means eventually, the full-length content is a means to an end. If clips really are the present and future of media and reach online, one begins to wonder what justifies making the unclipped, complete content in the first place.
Now that the clipping cat is fully out of the bag, companies offering the service will likely be busy; maybe some firm really will hire a “Chief Clipping Officer” (though I’d advise against it). It might be effective in pumping views for a while — but if we are to believe the platforms where this kind of recycled content lives, reused clips may have a short shelf life online. Meta has said it’s cracking down on “unoriginal” content that includes many of the hallmarks of clippers: “adding borders, inserting captions, and changing the reel’s speed” are specifically called out. Clipping companies are pulling in millions of dollars by condensing politicians, podcasters, indie rock bands, and Silicon Valley technocrats into bite-sized content — but even the clippers need something to clip from. If all that matters is going viral, the value of producing anything more complete begins to diminish, and so does the viewers’ incentive to engage with anything beyond the clips.
In April, a clip of Tucker Carlson’s podcast circulated online, in which the right wing media personality said he would be “tormented” for playing a role in Donald Trump getting elected. The excerpt was shared by TMZ, discussed on The View, and clipped by Headquarters, the social media account run by former Kamala Harris staffers. What was part of the episode but that didn’t hit the clip farms was Carlson saying Barack Obama “hated” white people, or brushing off criticism that Trump is racist because he was known to be romantically involved with “everyone.” Carlson’s repentance for supporting Trump is a perfectly executed soundbite, just the right size to satiate a viewer who can then scroll to the next thing. The clip becomes more urgent than the thing it existed to promote.
Oil prices were sharply lower on Wednesday, extending losses after Axios reported that the U.S. and Iran were closing in on an agreement to bring an end to the conflict.
International benchmark Brent crude futures slipped 8.2% to $100.83 per barrel as of 6:43 a.m. ET, while U.S. West Texas Intermediate futures lost 9.8% to $92.28. Both oil contracts settled more than 3.9% lower in the previous session.
Axios reported, citing two U.S. officials and two other sources briefed on the issue, that the White House believes it is nearing a one-page, 14-point memorandum of understanding to end the war and establish a framework for more detailed nuclear talks.
The U.S. reportedly expects Iran to respond on several key points over the next 48 hours. Nothing has been agreed yet between the two sides, but the sources told Axios that this was the closest Washington and Tehran had been to an agreement since the war began on Feb. 28.
A spokesperson for Iran’s foreign ministry told CNBC they were “evaluating” Washington’s 14-point peace proposal. The White House, meanwhile, was not immediately available to comment.
Iran said earlier on Wednesday that it would only accept a peace deal that was “fair.”
U.S. President Donald Trump on Tuesday announced in a Truth Social post that the U.S. would temporarily halt “Project Freedom,” a military effort launched just a day earlier to escort commercial vessels through the Strait of Hormuz, citing progress in negotiations with Iran toward a final agreement.
The Trump administration said roughly 23,000 seafarers across vessels from 87 countries have been stranded in the Persian Gulf following Iran’s effective shutdown of the strait.
“A deal that normalises oil flows through the Strait of Hormuz is crucial,” Warren Patterson, head of commodities strategy at Dutch bank ING, said in a research note.
“Roughly 13 mb/d of disrupted supply is being largely offset by inventory, which is clearly declining rapidly. This leaves the market more vulnerable with each passing day. Tighter stocks will only leave the oil market trading in an ever more volatile manner,” he added.
Surging oil and energy costs were already creating demand destruction globally, Azimut Group’s co-head of fixed income Nicolo Bocchin warned, adding that even if the waterway reopens, normalization in shipping and trade flows would still take “weeks and weeks.”
We have not been to fully verify each account but where possible, we have sought to corroborate stories through photos, location data and online records. Ludmilla’s own home was destroyed, she says, when the Kakhovka Dam further up the Dnipro river was blown up under Russian occupation in June 2023, causing catastrophic flooding.
The entries cited during the trial were written between 2015, when OpenAI was founded, and 2023, when Brockman and OpenAI CEO Sam Altman were briefly ousted as leaders over the OpenAI board’s alleged safety concerns.
Musk hopes the diary entries paint Brockman as a money-hungry executive who, early on, cared little about OpenAI’s mission.
To overcome that characterization of his mindset in OpenAI’s early days, Brockman has the challenging task of convincing the court that, instead, they show the opposite: displaying the careful musings of the person who is perhaps most committed to OpenAI’s mission.
Musk’s attorney, Steven Molo, spent the first day of Brockman’s testimony isolating passages and demanding that Brockman answer for the apparent greed that his journal entries revealed.
For example, Brockman drafted a journal in 2017, around the same time he testified that Musk had delivered an ultimatum: Either Musk would have full control over a for-profit arm of OpenAI, or OpenAI would remain a nonprofit.
In that entry, Brockman appears greedy, writing that “we’ve been thinking that maybe we should just flip to a for-profit. Making the money for us sounds great and all.”
And Brockman, of course, did make a lot of money after OpenAI created a for-profit arm in 2018, with his stake today worth about $30 billion. More than a dozen times, NBC News reported, Molo asked Brockman to justify his stake, while repeatedly pointing to the journal entry in which the OpenAI president also said that $1 billion was all he wanted for his career goal.
“Financially, what will take me to $1B?” Brockman wrote in 2017, while mulling whether Musk was the “glorious leader” he wanted to run OpenAI or if he should back Altman.
At a contentious point, Molo asked whether Brockman would consider giving $29 billion back to the nonprofit arm. But Brockman said no, pointing out that he received the stake well before ChatGPT’s release spiked OpenAI’s value. He also emphasized that he helped grow the best-funded nonprofit in the world. According to The Information, Molo then likened Brockman to a “bank robber” who downplays the theft of only $1 million because there’s much more money left in the bank.
‘We want to be clear: this is not facial recognition,’ Facebook says.
‘We want to be clear: this is not facial recognition,’ Facebook says.
Facebook and Instagram have a new way to detect and remove users under 13: AI bone structure analysis. In a blog post on Tuesday, Meta — Facebook and Instagram’s parent company — says its AI system will scan photos and videos posted to its platforms for “general themes and visual cues,” including height and bone structure.
“We want to be clear: this is not facial recognition,” Meta says in the blog post, adding that it “does not identify the specific person in the image.” This system is part of Meta’s efforts to keep kids under 13 off its platforms, and will also analyze posts, comments, bios, and captions to search for “contextual clues” that someone might be underage.
Meta’s AI-powered facial analysis, which is only available in “select” countries including the US ahead of a wider rollout, seems similar to the face-scanning tech offered by age verification services like Yoti and k-ID. Facebook and Instagram will deactivate accounts identified as underage, and the owner will need to verify their age to prevent it from deletion.
The announcement comes just days after a New Mexico jury found that Meta violated state law by misleading customers about the safety of its platforms and failing to protect children from child predators. Meta must pay $375 million as a result, and may have to implement changes that the company has already threatened to leave the state over.
Separately, Meta is expanding the technology it uses on Instagram to automatically identify and place users between 13 and 18 into Teen Accounts. These accounts come with stricter content controls, block messages from strangers, and prevent users under 16 from livestreaming. Instagram rolled out the tech in 2024, and now Facebook will do the same for users in the US, followed by a rollout in the UK and EU in June.
In its announcement, Meta continues to advocate for age verification at the app store and operating system level, an approach that’s gaining traction in Congress and some states, including California and Colorado.
Up until the weekend, global markets had been betting on a fragile ceasefire between the U.S. and Iran turning into a longer-term peace deal.
But escalatory rhetoric, action over the Strait of Hormuz, and fresh Iranian attacks on the United Arab Emirates over the past 48 hours, have led experts to warn war could be back.
Market analysts said the latest developments could mark an inflection point in the war and a critical moment for financial markets and global energy supplies, which are dwindling as the Strait of Hormuz remains largely closed.
“It’s an incredibly delicate moment,” Ben Powell, chief investment strategist for APAC at BlackRock, told CNBC Tuesday.
“It’s very unsettling to have our first missile warnings here in Abu Dhabi for several weeks. We were all hoping that was behind us,” the UAE-based strategist said.
“Looking forward, I think there is a genuine complexity as to whether this escalation yesterday was just part of the negotiation — Iran showing that they’ve still got cards to play, perhaps signaling to the UAE that the UAE can leave the OPEC, but energy leaving the region is still dependent on the favor from Iran — or it could be a beginning of a more difficult moment,” he told CNBC’s “Access Middle East.”
Energy and other key parts of the global economy aren’t flowing, he said, adding that it now feels like we’re approaching a “critical moment” where inventories have been run down, and the delayed impact of the energy shock is starting to come into sharper focus.
Global markets were edgy Tuesday morning after the latest developments in the Middle East at the weekend, which saw the U.S. try to end a stalemate over the blockaded Strait of Hormuz by launching “Project Freedom,” essentially, a bid to “free” stranded ships in the strait and to safely escort them out of the waterway, which has been blockaded by both Iran and the U.S.
Those attempts were met with Iranian resistance, with the U.S. saying it had sunk several Iranian boats as a result of skirmishes in the channel, although Iran denied any of its boats had been sunk.
Iran later attacked oil infrastructure in the UAE with missiles and drones, seemingly resuming a strategy of lashing out against its Middle East neighbors to get them to pressure the U.S. to end the war.
Financial markets were a mixed bag on Tuesday, with bourses lower in Asia and mixed in Europe, while U.S. stock futures moved higher ahead of the forthcoming session on Wall Street.
That apparent bounceback in the U.S. comes after the major averages suffered declines on Monday and oil prices rose amid growing concerns that the conflict in the Middle East could resume with a vengeance.
Markets are likely to remain on tenterhooks as they wait to see what happens next, with geopolitical watchers now warning that a complete collapse of the already tentative U.S.-Iran ceasefire is looking increasingly likely.
“The question of the week is whether geopolitical risks will remain messy but contained, or break through to weigh on markets and corporate earnings,” Tina Fordham, founder of Fordham Global Foresight, commented in analysis Tuesday.
“Judging by the apparent resumption today of Iranian attacks on targets in the UAE itself and on ships attempting to pass through the Strait of Hormuz, I’m currently leaning toward the latter,” she warned.
Describing the apparent resumption of Iranian drone and missile attacks against the UAE as “the biggest escalation in some weeks,” she said Tehran had shown the U.S. that it still has fight left in it, peace deal or none.
“Iran is signalling that they still have the capacity to inflict pain and won’t be forced into capitulation. The U.S. increasingly faces a choice between a long war it doesn’t want to fight, or a bad, embarrassing deal,” Fordham noted in emailed comments.
The latest turn in events in the Iran war appears to have been precipitated by the U.S.’ ‘Project Freedom’ strategy, announced on Sunday, aimed at forcing a reopening of the Strait of Hormuz. The move came after the White House appeared to have grown frustrated at an impasse over a peace deal with Iran. Pakistan-mediated talks have stalled in recent weeks, and both sides are at odds over an agreement.
U.S. President Donald Trump told Fox News Monday that Iran would be “blown off the face of the earth” if it targeted U.S. ships that are protecting commercial vessels transiting the strait.
Tehran signaled overnight that it is still interested in Pakistan-mediated peace talks with the U.S. and warned that a political solution was the only way to break the deadlock over a deal.
“As talks are making progress with Pakistan’s gracious effort, the U.S. should be wary of being dragged back into quagmire by ill-wishers. So should the UAE. Project Freedom is Project Deadlock,” Iran’s Foreign Minister Abbas Araghchi commented on X overnight.
He said the latest developments in the war “make clear that there’s no military solution to a political crisis.”
Every few centuries, changes in how information moves reshape how societies govern themselves. The printing press spread vernacular literacy, helping give rise to the Reformation and, eventually, representative government. The telegraph made it possible to administer vast nations like the US, accelerating the growth of the modern bureaucratic state. Broadcast media created shared national audiences, which in turn fueled mass democracy.
We are now in the early stages of another such shift. Faster than many realize, AI is becoming the primary interface through which we form beliefs and participate in democratic self-governance. If left unchecked, this shift could further strain America’s already fragile institutions. But it could also help address long-standing problems, like lagging civic engagement and deepening polarization. What happens next depends on design choices that are already being made, whether we know it or not.
Start with what might be called the epistemic layer—how we come to know things. People are increasingly relying on AI to know what is true, what is happening, and whom to trust. Search is already substantially AI-mediated. The next generation of AI assistants will synthesize information, frame it, and present it with authority. For a growing number of people, asking an AI will become the default way to form views on a candidate, a policy, or a public figure. Whoever controls what these models say therefore has increasing influence over what people believe.
Technology has always shaped the way citizens interact with information. But a new problem will soon arise in the form of personal AI agents, which can change not only how people receive information but how they act on it. These systems will conduct research, draft communications, highlight causes, and lobby on a user’s behalf. They will inform decisions such as how to vote on a ballot measure, which organizations are worth supporting, or how to respond to a government notice. They will, in a meaningful sense, begin to mediate the relationship between individuals and the institutions that govern them.
We’ve already seen with social media what happens when algorithms optimize for engagement over understanding. Platforms do not need to have an explicit political agenda to produce polarization and radicalization. An agent that knows your preferences and your anxieties—one shaped to keep you engaged—poses the same risks. And in this case the risks may be even more difficult to detect, because an agent presents itself as your advocate. It speaks for you, acts on your behalf, and may earn trust precisely through that intimacy.
Now zoom out to the collective. AI agents and humans could soon participate in the same forums, where it may be impossible to tell them apart. Even if every individual AI agent were well-designed and aligned with its user’s interests, the interactions of millions of agents could produce outcomes that no individual wanted or chose. For example, research shows that agents displaying no individual bias can still generate collective biases at scale. And setting aside what agents do to each other, there is what they do for their users. A public sphere in which everyone has a personalized agent attuned to their existing views is not, in aggregate, a public sphere at all. It is a collection of private worlds, each internally coherent but collectively inhospitable to the kind of shared deliberation that democracy requires.
Taken together, these three transformations—in how we know, how we act, and how we engage in collective governance—amount to a fundamental change in the texture of citizenship. In the near future, people will form their political views through AI filters, exercise their civic agency through AI agents, and participate in institutions and public discussions that are themselves shaped by the interactions of millions of such agents.
Today’s democracy is not ready for this. Our institutions were designed for a world in which power was exercised visibly, information traveled slowly enough to be contested, and reality felt more shared, if imperfectly. All of this was already fraying long before generative AI arrived. And yet this need not be a story of decline. Avoiding that outcome requires us to design for something better.
On the informational layer, AI companies must ramp up existing efforts to ensure that models’ outputs are truthful. They should also explore some promising early findings that AI models can help reduce polarization. A recent field evaluation of AI-generated fact checks on X found that people with a variety of political viewpoints deemed AI-written notes more helpful than human-written ones. The paper is yet to be peer-reviewed, but that is a potentially revolutionary finding: AI-assisted fact-checking may be able to achieve the kind of cross-partisan credibility that has eluded most manual human efforts. Greater understanding of and transparency about how models make these assertions and prioritize sources in the process could help build further public trust.
On the agentic layer, we need ways to evaluate whether AI agents faithfully represent their users. An agent must never have an agenda of its own or misrepresent its user’s views—a technically daunting requirement in domains where users may have not explicitly stated any preferences. But faithful representation also cannot become an accessory to motivated reasoning. An agent that refuses to present uncomfortable information, that shields its user from ever questioning prior beliefs or fails to adjust to a change of heart, is not acting in the person’s best interest.
Finally, on the institutional level, policymakers should hurry to harness AI’s potential to make governance more responsive and legitimate. Several states and localities are already using AI-mediated platforms to conduct democratic deliberation at scale, building on research showing that AI mediators can help citizens find common ground. As agents become increasingly common participants in public input processes—and there is already evidence that bots are skewing those processes—identity verification for both humans and their agentic proxies must be built in from the start.
What is needed is a new generation of democratic infrastructure, technological and institutional, built for the world that is actually here. Failing to design for democratic outcomes, in a domain this consequential, means designing for something else. And the history of unaccountable power does not leave much room for optimism about what that something else tends to be.
Andrew Sorota and Josh Hendler lead work on AI and democracy at the Office of Eric Schmidt.
Analysts say he is entering a landscape still dominated by the DMK and the AIADMK – a duopoly that mostly appears stable on paper, yet shows signs of fatigue on the ground. That, they argue, is opening space for new political experiments – and for figures like Vijay to test how far star power can translate into durable political authority.
