The use case for AI video is to churn out clips of existing TV and movie characters for your amusement. [THR]

Hollywood got wind of OpenAI’s Sora 2. Agencies called OpenAI to see what the deal was. [THR]

OpenAI told them whatever they wanted to hear:

repeatedly talking up an opt-in regime that would protect the agency’s clients against the misuse of their intellectual property and likenesses.

But when the agents talked to each other:

We started exchanging notes with others having similar conversations and realized we’re all hearing different things.

OpenAI told some agents they would never use actor likenesses without opt-in! Others were told Sora would be opt-out.

The WME agency was told each actor would have to opt out individually. Then, on 29 September, WME was told Sora would not use their clients’ faces or voices without opt-in permission.

Sora went live 30 September, using all the faces and voices anyway. OpenAI announced an opt-out regime, opting out one property at a time.

That’s not how copyright or personality rights work. There’s a century of precedent — you do a deal and you pay. Anthropic successfully claimed fair use on training — but OpenAI has built a machine to churn out violations.

WME has notified OpenAI it does not have permission to use any of WME’s clients. Disney has notified OpenAI that it could just sue under copyright if OpenAI does not desist. [THR]

OpenAI’s going ahead full-steam anyway. Because OpenAI has to get money in, or it dies.

• Video — Podcast

Click here to go see the bonus panel!

Hovertext:
Works especially well for telling people they're not in irrirrerror.

OpenAI has been flat broke its whole existence. In late 2025, it’s even more flat broke than that. Sam’s been frantically lining up about a trillion dollars’ worth of deals. Well, deals to make a deal.

In the first half of 2025, OpenAI had $4.3 billion in revenue! — with a net loss of $13.5 billion. Their “annualised” revenue is $13 billion of late — which means they took their very best month and multiplied it by twelve. But subscription numbers still look flat. [Register; FT, archive]

A lot of that $13.5 billion loss is “remeasurement of convertible interest rights,” which seems to mean pricing down the share price of OpenAI. Well, the alleged value of a share of future revenue from the for-profit subsidiary of the OpenAI nonprofit, if it ever makes a penny.

The press talks about this like it’s shares of stock, and it sort of is, if you stretch? But it’s different in ways that are important here.

SoftBank can put these OpenAI pseudo-shares on its books and watch the SoftBank share price go up. But these “interest rights” don’t work like conventional stock, or even conventional private company equity. The price of these pseudo-shares is a function of the vibe around how fast-talking Sam Altman is today.

“Remeasurement of convertible interest rights” means Sam’s had to adjust his fast talking to make a few less ludicrous promises about how much money OpenAI’s totally going to make.

But at some point, the rubber hits the road. OpenAI has to get some actual income. One source admitted to the Financial Times that only 5% of ChatGPT users are actually paying customers.

The bot has various sales features built in through Shopify, and OpenAI takes a cut from those. Altman is also talking about running personalised advertising. Just imagine the advertising chats!

Deutsche Bank says ChatGPT subscriptions in Europe have “flatlined in the major European markets over the past four months”. [Fortune]

Sam needs new markets, and he needs them fast. So he’s offering to relax the restrictions on GPT-5 about emotional chat, and making the bot more adult: [Twitter]

In a few weeks, we plan to put out a new version of ChatGPT that allows people to have a personality that behaves more like what people liked about 4o (we hope it will be better!).

… In December, as we roll out age-gating more fully and as part of our “treat adult users like adults” principle, we will allow even more, like erotica for verified adults.

The whole Internet saw that last sentence and shuddered at the concept of GPT porn bots.

Sam tried to walk it back — it “blew up on the erotica point much more than I thought it was going to!” Yeah, sure. [Twitter]

But I’m not inclined to put much faith in Sam on this one. Because Elon Musk’s Grok is going hard into the sexy chat “companions,” starting with his “Ani” bot, which is short for “anime” ’cos it’s really obviously Misa from Death Note. [Verge]

The sort of people who pay for X The Everything App are just the customers for that bot. And, of course, for Grok Spicy Mode. There’s a market, and Sam Altman’s going for it!

OpenAI is still the flagship and linchpin of the AI bubble, which is holding up the whole US stock market. As Bluesky user Jack Klompus put it: [Bluesky]

The entire economy now rests on a porn pivot by a piracy company.

• Video — Podcast

It's my 61st birthday this weekend and I have to say, I never expected to get to be this old—or this weirded-out by the world I'm living in, which increasingly resembles the backstory from a dystopian 1970s SF novel in which two-fisted billionaires colonize space in order to get away from the degenerate second-hander rabble downstairs who want to survive their John W. Campbell-allocated banquet of natural disasters. (Here's looking at you, Ben Bova.)

Notwithstanding the world being on fire, an ongoing global pandemic vascular disease that is being systematically ignored by governments, Nazis popping out of the woodwork everywhere, actual no-shit fractional trillionaires trying to colonize space in order to secede from the rest of the human species, an ongoing European war that keeps threatening to drag NATO into conflict with the rotting zombie core of the former USSR, and an impending bubble collapse that's going to make 2000 and 2008 look like storms in a teacup ...

I'm calling this the pivotal year of our times, just as 1968 was the pivotal year of the post-1945 system, for a number of reasons.

It's pretty clear now that a lot of the unrest we're seeing—and the insecurity-induced radicalization—is due to an unprecedented civilizational energy transition that looks to be more or less irreversible at this point.

Until approximately 1750, humanity's energy budget was constrained by the available sources: muscle power, wind power (via sails and windmills), some water power (via water wheels), and only heat from burning wood and coal (and a little whale oil for lighting).

During the 19th century we learned to use combustion engines to provide motive power for both stationary machines and propulsion. This included powering forced ventilation for blast furnaces and other industrial processes, and pumps for water and other working fluids. We learned to reform gas from coal for municipal lighting ("town gas") and, later, to power dynamos for municipal electricity generation. Late in the 19th century we began to switch from coal (cumbersome, bulky, contained non-combustible inclusions) to burning fractionated oil for processes that demanded higher energy densities. And that's where we stuck for most of the long 20th century.

During the 20th century, the difficulty of supporting long-range military operations led to a switch from coal to oil—the pivotal event was the ultimately-disastrous voyage of the Russian Baltic fleet to the Sea of Japan in 1906, during the Russo-Japanese war. From the 1890s onwards Russia had been expanding into Siberia and then encroaching on the edges of the rapidly-weakening Chinese empire. This brought Russia into direct conflict with Japan over Korea (Japan, too, had imperial ambitions), leading to the outbreak of war in 1905—when Japan wiped out the Russian far-eastern fleet in a surprise attack. (Pearl Harbor in 1941 was not that surprising to anyone familiar with Japanese military history!) So the Russian navy sent Admiral Zinovy Rozhestvensky, commander of the Baltic Fleet, to the far east with the hastily-renamed Second Pacific Squadron, whereupon they were sunk at the Battle of Tsushima.

Rozhestvensky had sailed his fleet over 18,000 nautical miles (33,000 km) from the Baltic Sea, taking seven months and refueling numerous times at sea with coal (around a quarter of a million tons of it!) because he'd ticked off the British and most ports were closed to him. To the admiralties watching from around the world, the message was glaringly obvious—coal was a logistical pain in the arse—and oil far preferable for refueling battleships, submarines, and land vehicles far from home. (HMS Dreadnought, the first turbine-powered all-big-gun battleship, launched in 1905, was a transitional stage that still relied on coal but carried a large quantity of fuel oil to spray on the coal to increase its burn rate: later in the decade, the RN moved to oil-only fueled warships.)

Spot the reason why the British Empire got heavily involved in Iran, with geopolitical consequences that are still playing out to this day! (The USA inherited large chunks of the British empire in the wake of the second world war: the dysfunctional politics of oil are in large part the legacy of applying an imperial resource extraction model to an energy source.)

Anyway. The 20th century left us with three obvious problems: automobile driven suburban sprawl and transport infrastructure, violent dissatisfaction among the people of colonized oil-producing nations, and a massive burp of carbon dioxide emissions that is destabilizing our climate.

Photovoltaic cells go back to 1839, but until the 21st century they remained a solution in search of very specific problems: they were heavy, produced relatively little power, and degraded over time if left exposed to the sun. Early PV cells were mainly used to provide power to expensive devices in inaccessible locations, such as aboard satellites and space probes: it cost $96 per watt for a solar module in the mid-1970s. But we've been on an exponential decreasing cost curve since then, reaching $0.62/watt by the end of 2012, and it's still on-going.

China is currently embarked on a dash for solar power which really demands the adjective "science-fictional", having installed 198GW of cells between January and May, with 93GW coming online in May alone: China set goals for reaching net-zero carbon emissions by 2030 in 2019 and met their 2030 goal in 2024, so fast is their transition going. They've also acquired a near-monopoly on the export of PV panels because this roll-out is happening on the back of massive thin-film manufacturing capacity.

The EU also hit a landmark in 2025, with more than 50% of its electricity coming from renewables by late summer. It was going to happen sooner or later, but Russia's attack on Ukraine in 2022 sped everything up: Europe had been relying on Russian exports of natural gas via the Nordstream 1 and 2 pipelines, but Russia—which is primarily a natural resource extraction economy—suddenly turned out to be an actively hostile neighbour. (Secondary lesson of this war: nations run by a dictator are subject to erratic foreign policy turns—nobody mention Donald Trump, okay?) Nobody west of Ukraine wanted to be vulnerable to energy price warfare as a prelude to actual fighting, and PV cells are now so cheap that it's cheaper to install them than it is to continue mining coal to feed into existing coal-fired power stations.

This has not gone unnoticed by the fossil fuel industry, which is collectively shitting itself. After a couple of centuries of prospecting we know pretty much where all the oil, coal, and gas reserves are buried in the ground. (Another hint about Ukraine: Ukraine is sitting on top of over 670 billion cubic metres of natural gas: to the dictator of a neighbouring resource-extraction economy this must have been quite a draw.) The constant propaganda and astroturfed campaigns advocating against belief in climate change must be viewed in this light: by 2040 at the latest, those coal, gas, and oil land rights must be regarded as stranded assets that can't be monetized, and the land rights probably have a book value measured in trillions of dollars.

China is also banking on the global shift to transport using EVs. High speed rail is almost always electrified (not having to ship an enormous mass of heavy fuel around helps), electric cars are now more convenient than internal combustion ones to people who live in dense population areas, and e-bikes don't need advocacy any more (although roads and infrastructure friendly to non-motorists—pedestrians and public transport as well as cyclists—is another matter).

Some forms of transport can't obviously be electrified. High capacity/long range aviation is one—airliners get lighter as they fly because they're burning off fuel. A hypothetical battery powered airliner can't get lighter in flight: it's stuck with the dead weight of depleted cells. (There are some niches for battery powered aircraft, including short range/low payload stuff, air taxis, and STOVL, but they're not going to replace the big Airbus and Boeing fleets any time soon.)

Some forms of transport will become obsolescent in the wake of a switch to EVs. About half the fossil fuel powered commercial shipping in use today is used to move fossil fuels around. We're going to be using crude oil for the foreseeable future, as feedstock for the chemical and plastics industries, but they account for a tiny fraction of the oil we burn for transport, including shipping. (Plastic recycling is over-hyped but might eventually get us out of this dependency—if we ever get it to work efficiently.)

So we're going through an energy transition period unlike anything since the 1830s or 1920s and it's having some non-obvious but very important political consequences, from bribery and corruption all the way up to open warfare.

The geopolitics of the post-oil age is going to be interestingly different.

I was wrong repeatedly in the past decade when I speculated that you can't ship renewable electricity around like gasoline, and that it would mostly be tropical/equatorial nations who benefited from it. When Germany is installing rooftop solar effectively enough to displace coal generation, that's a sign that PV panels have become implausibly cheap. We have cars and trucks with reasonably long ranges, and fast-charger systems that can take a car from 20% to 80% battery capacity in a quarter of an hour. If you can do that to a car or a truck you can probably do it to a tank or an infantry fighting vehicle, insofar as they remain relevant. We can do battery-to-battery recharging (anyone with a USB power bank for their mobile phone already knows this) and in any case the whole future of warfare (or geopolitics by other means) is up in the air right now—quite literally, with the lightning-fast evolution of drone warfare over the past three years.

The real difference is likely to be that energy production is widely distributed rather than concentrated in resource extraction economies and power stations. It turns out that PV panels are a great way of making use of agriculturally useless land, and also coexist well with some agricultural practices. Livestock likes shade and shelter (especially in hot weather) so PV panels on raised stands or fences can work well with sheep or cattle, and mixed-crop agriculture where low-growing plants are sheltered from direct sunlight by taller crops can also work with PV panels instead of the higher-growing plants. You can even in principle use the power from the farm PV panels to drive equipment in greenhouses: carbon dioxide concentrators, humidifiers, heat pumps to prevent overheating/freezing, drainage pumps, and grow lamps to drive the light-dependent reactions in photosynthesis.

All of which we're really going to need because we've passed the threshold for +1.5 °C climate change, which means an increasing number of days per year when things get too hot for photosynthesis under regular conditions. There are three main pathways for photosynthesis, but none of them deal really well with high temperatures, although some adaptation is possible. Active cooling is probably impractical in open field agriculture, but in intensive indoor farming it might be an option. And then there's the parallel work on improving how photosynthesis works: an alternative pathway to the Calvin cycle is possible and the enzymes to make it work have been engineered into Arabidopsis, with promising results.

In addition to the too-many-hot-days problem, climate change means fluctuations in weather: too much wind, too much rain—or too little of both—at short notice, which can be physically devastating for crops. Our existing staple crops require a stable, predictable climate. If we lose that, we're going to have crop failures and famines by and by, where it's not already happening. The UK has experienced three of its worst harvests in the past century in this decade (and this decade is only half over). As long as we have global supply chains and bulk shipping we can shuffle food around the globe to cover localized shortfalls, but if we lose stable agriculture globally for any length of time then we are all going to die: our economic system has shifted to just-in-time over the past fifty years, and while it's great for efficiency, efficiency is the reciprocal of resilience. We don't have the reserves we would need to survive the coming turbulence by traditional means.

This, in part, explains the polycrisis: nobody can fix what's wrong using existing tools. Consequently many people think that what's going wrong can't be fixed. The existing wealthy elites (who have only grown increasingly wealthy over the past half century) derive their status and lifestyle from the perpetuation of the pre-existing system. But as economist Herbert Stein observed (of an economic process) in 1985, "if it can't go on forever it will stop". The fossil fuel energy economy is stopping right now—we've probably already passed peak oil and probably peak carbon: the trend is now inexorably downwards, either voluntarily into a net-zero/renewables future, or involuntarily into catastrophe. And the involuntary option is easier for the incumbents to deal with, both in terms of workload (do nothing, right up until we hit the buffers) and emotionally (it requires no sacrifice of comfort, of status, or of relative position). Clever oligarchs would have gotten ahead of the curve and invested heavily in renewables but the evidence of our eyes (and the supremacy of Chinese PV manufacturers in the global market) says that they're not that smart.

The traditional ruling hierarchy in the west had a major shake-up in 1914-19 (understatement: most of the monarchies collapsed) in the wake of the convulsion of the first world war. The elites tried to regain a degree of control, but largely failed due to the unstable conditions produced by the great depression and then the second world war (itself an emergent side-effect of fascist regimes' attempts to impose imperial colonial policies on their immediate neighbours, rather than keeping the jackboots and whips at a comfortable remove). Reconstruction after WW2 and a general post-depression consensus that emerged around accepting the lesser evil of social democracy as a viable prophylactic to the devil of communism kept the oligarchs down for another couple of decades, but actually-existing capitalism in the west stopped being about wealth creation (if it ever had been) some time in the 1960s, and switched gear to wealth concentration (the "he who dies with the most toys, wins" model of life). By the end of the 1970s, with the rise of Thatcherism and Reaganomics, the traditional wealthy elites began to reassert control, citing the spurious intellectual masturbation of neoliberal economics as justification for greed and repression.

But neoliberalism was repurposed within a couple of decades as a stalking-horse for asset-stripping, in which the state was hollowed out and its functions outsourced to the private sector—to organizations owned by the existing elites, which turned the public purse into a source of private profit. And we're now a couple of generations into this process, and our current rulers don't remember a time when things were different. So they have no idea how to adapt to a changing world.

Cory Doctorow has named the prevailing model of capitalist exploitation enshittification. We no longer buy goods, we buy services (streaming video instead of owning DVDs or tapes, web services instead of owning software, renting instead of buying), and having been captured by the platforms we rent from, we are then subject to rent extraction: the service quality is degraded, the price is jacked up, and there's nowhere to go because the big platforms have driven their rivals into bankruptcy or irrelevance:

It's a three stage process: First, platforms are good to their users; then they abuse their users to make things better for their business customers; finally, they abuse those business customers to claw back all the value for themselves. Then, they die.

This model of doing business (badly) is a natural consequence of the bigger framework of neoliberalism, under which a corporation's directors overriding duty is to maximize shareholder value in the current quarter, with no heed to the second and subsequent quarters hence: the future is irrelevant, feed me shouts the Audrey II of shareholder activism. Business logic has no room for the broader goals of maintaining a sustainable biosphere, or even a sustainable economy. And so the agents of business-as-usual, or Crapitalism as I call it, are at best trapped in an Abilene paradox in which they assume everyone else around them wants to keep the current system going, or they actually are as disconnected from reality as Peter Thiel (who apparently believes Greta Thunberg is the AntiChrist.)

if it can't go on forever it will stop

What we're seeing right now is the fossil fuel energy economy stopping. We need it to stop; if it doesn't stop, we're all going to starve to death within a generation or so. It's already leading to resource wars, famines, political upheaval, and insecurity (and when people feel insecure, they rally to demagogues who promise them easy fixes: hence the outbreaks of fascism). The ultra-rich don't want it to stop because they can't conceive of a future in which it stops and they retain their supremacy. (Also, they're children of privilege and most of them are not terribly bright, much less imaginative—as witness how easily they're robbed blind by grifters like Bernie Madoff, Sam Bankman Fried, and arguably Sam Altman). Those of them whose wealth is based in ownership of fossil fuel assets still in the ground have good reason to be scared: these are very nearly stranded assets already, and we're heading for a future in which electricity is almost too cheap to meter.

All of this is without tackling the other elephant in the room, which is the end of Moore's Law. Moore's Law has been on its death bed for over a decade now. We're seeing only limited improvements in computing and storage performance, mainly from parallelism. Aside from a very few tech bubbles which soak up all available processing power, belch, and ask for more, the all you can eat buffet for tech investors is over. (And those bubbles are only continuing as long as scientifically naive investors keep throwing more money at them.)

The engine that powered the tech venture capital culture (and the private equity system battening on it) is sputtering and dying. Massive AI data centres won't keep the coal mines running or the nuclear reactors building out (it's one of those goddamn bubbles: to the limited extent that LLMs are useful, we'll inevitably see a shift towards using pre-trained models running on local hardware). They're the 2025 equivalent of 2020's Bored Ape NFTs (remember those?). The forecast boom in small modular nuclear reactors is going to fizzle in the face of massive build-out of distributed, wildly cheap photovoltaic power plus battery backup. Quantum computing isn't going to save the tech sector, and that's the "next big thing" the bubble-hypemongers have been saving for later for the past two decades. (Get back to me when you've got hardware that can factor an integer greater than 31.)

If we can just get through the rest of this decade without widespread agricultural collapses, a nuclear war, a global fascist international dictatorship taking hold, and a complete collapse of the international financial system caused by black gold suddenly turning out to be worthless, we might be pretty well set to handle the challenges of the 2030s.

But this year, 2025, is the pivot. This can't go on. So it's going to stop. And then—

In 2022, Jason Allen outraged artists around the world when he won the Colorado State Fair Fine Arts Competition with a piece of AI-generated art. A month later, he tried to copyright the pictures, got denied, and started a fight with the U.S. Copyright Office (USCO) that dragged on for three years. In August, he filed a new brief he hopes will finally give him a copyright over the image Midjourney made for him, called Théâtre D’opéra Spatial. He’s also set to start selling oil-print reproductions of the image. 

A press release announcing both the filing and the sale claims these prints “[evoke] the unmistakable gravitas of a hand-painted masterwork one might find in a 19th-century oil painting.” The court filing is also defensive of Allen’s work. “It would be impossible to describe the Work as ‘garden variety’—the Work literally won a state art competition,” it said. 

“So many have said I’m not an artist and this isn’t art,” Allen said in a press release announcing both the oil-print sales and the court filing. “Being called an artist or not doesn’t concern me, but the work and my expression of it do. I asked myself, what could make this undeniably art? What if I could create Théâtre D’opéra Spatial as if it were physically created by hand? Not actually, of course, but what if I could achieve that using technology? Surely that would be the answer.”

Allen’s 2022 win at the Colorado State Fair was an inflection point. The beta version for the image generation software Midjourney had launched a few months before the competition and AI-generated images were still a novelty. We were years away from the nightmarish tide of slop we all live with today, but the piece was highly controversial and represented one of the first major incursions of AI-generated work into human spaces.

Théâtre D’opéra Spatial was big news at the time. It shook artistic communities and people began to speak out against AI-generated art. Many learned that their works had been fed into the training data for these massive data hungry art generators like Midjourney. About a month after he won the competition and courted controversy, Allen applied for a copyright of the image. The USCO rejected it. He’s been filing appeals ever since and has thus far lost every one.

The oil-prints represent an attempt to will the AI-generated image into a physical form called an “elegraph.” These won’t be hand painted versions of the picture Midjourney made. Instead, they’ll employ a 3D printing technique that uses oil paints to create a reproduction of the image as if a human being made it, complete—Allen claimed—with brushstrokes. 

“People said anyone could copy my work online, sell it, and I would have no recourse. They’re not technically wrong,” Allen said in the press release. “If we win my case, copyright will apply retroactively. Regardless, they’ll never reproduce the elegraph. This artifact is singular. It’s real. It’s the answer to the petulant idea that this isn’t art. Long live Art 2.0.”

The elegraph is the work of a company called Arius which is most famous for working with museums to conduct high quality scans of real paintings that capture the individual brushstrokes of masterworks. According to Allen’s press release, Arius’ elegraphs of Théâtre D’opéra Spatial will make the image appear as if it is a hand painted piece of art through “a proprietary technique that translates digital creation into a physical artifact indistinguishable in presence and depth from the great oil paintings of history…its textures, lighting, brushwork, and composition, all recalling the timeless mastery of the European salons.”

Allen and his lawyers filed a request for a summary judgement with the U.S. District Court of Colorado on August 8, 2025. The 44 page legal argument rehashes many of the appeals and arguments Allen and his lawyers have made about the AI-generated image over the past few years.

“He created his image, in part, by providing hundreds of iterative text prompts to an artificial intelligence (“AI”)-based system called Midjourney to help express his intellectual vision,” it said. “Allen produced this artwork using ‘hundreds of iterations’ of prompts, and after he ‘experimented with over 600 prompts,’ he cropped and completed the final Work, touching it up manually and upscaling using additional software.”

Allen’s argument is that prompt engineering is an artistic process and even though a machine made the final image, he says he should be considered the artist because he told the machine what to do. “In the Board’s view, Mr. Allen’s actions as described do not make him the author of the Midjourney Image because his sole contribution to the Midjourney Image was inputting the text prompt that produced it,” a 2023 review of previous rejections by the USCO said.

During its various investigations into the case, the USCO did a lot of research into how Midjourney and other AI-image generators work. “It is the Office’s understanding that, because Midjourney does not treat text prompts as direct instructions, users may need to attempt hundreds of iterations before landing upon an image they find satisfactory. This appears to be the case for Mr. Allen, who experimented with over 600 prompts,” its 2023 review said.

This new filing is an attempt by Allen and his lawyers to get around these previous judgements and appeal to higher courts by accusing the USCO of usurping congressional authority.  “The filing argues that by attempting to redefine the term “author” (a power reserved to Congress) the Copyright Office has acted beyond its lawful authority, effectively placing itself above judicial and legislative oversight.”

We’ll see how well that plays in court. In the meantime, Allen is selling oil-prints of the image Midjourney made for him.