Why China’s AI Strategy Has The US Spooked; India 2026 H1 Box Office

Liz Shackleton

28 Jul 2026 — 10 min read

China and the US are considering restricting access to each other's AI models

We’re having another summer in which AI, war and wildfires are dominating the news cycle, while in India, students have been attacked with batons and tear gas as they protest for education reform. In the film and streaming industries, we’re being served up another dose of mega-mergers and media concentration, while governments rearrange the deckchairs on the Titanic in their attempts to shore up flagging theatrical and public broadcasting sectors. So in other words, it’s just another summer on Planet Earth.

In this edition of Streamlined we’re looking at AI – at least some of the more ironic aspects of the tech wars between the US and China – as well as taking the temperature of the Indian box office at the half year mark. In the next edition, we’ll be looking at how AI chatbots and Google’s AI Overviews are decimating the publishing industry (relevant if you need the press to help publicise your film) as well as reporting on box office in some other Asian territories.

Subscribe to Streamlined

US & China AI Battle: Who's Stealing From Who?

The creative industries are, of course, in the front line of those sectors that are being most impacted by the rapid escalation of AI technology, and while Streamlined has been planning to write about how it's being used in the worlds of film and streaming, at this precise moment it feels more important to write about the bigger picture in global tech.

We all know that (whether they admit to it or not) film and TV companies are now using AI tools across almost every stage of the production and distribution process – from script analysis and storyboarding, to editing, VFX, dubbing & subbing and creating marketing materials – while some filmmakers are going the whole hog and producing AI-generated video, some of which doesn’t involve human actors at all.

After watching a fair amount of AI-generated video over the past few months (some without human actors, some combining actors and AI-powered motion capture) my first thoughts have been, somewhat predictably, that AI should be used as a tool, not to create whole films and series. But if you insist on going down that route, the end result should be regarded as photorealistic animation, not live-action filmmaking, and there’s no point making it unless the storytelling and character development are also good.

My second thought is one that should become a mantra in today’s film and TV industries: “Just because you can, doesn’t mean you should”. More than ever I’m realising that great art comes from scarcity and struggle, not from limitless possibilities. Access to AI prompts doesn’t always equate to access to imagination and sound psychological processes, in fact quite the opposite. I’ve seen a lot of things in AI-generated videos that human beings don’t ever really need to see.

But AI is definitely not going anywhere – it’s not the latest buzzy thing like blockchain, Web3 and NFTs – that will be the subject of a thousand seminars before its slowly fades away. So it’s important to understand the business models behind it, who controls and can manipulate the technology, how it’s being regulated, and what that means for our personal privacy, career prospects and mental well-being. And after doing a whole lot of research, I have to say the current landscape is not looking good.

In the US, the dominant Large Language Models (LLMs) are being controlled by what has rapidly become a duopoly between OpenAI and Anthropic, who are pouring billions into building frontier models, artificially inflating stock markets (although a correction kicked off this week), and conveniently tossing aside their “kumbaya” terminology about benefitting humanity as they gear up for their initial public offerings (IPOs).

Right now, regulation is not really happening aside from the usual rhetoric about sanctions and trade tariffs, which should not surprise us while Congress keeps ceding so much power to Donald Trump. Indeed, recent news about the rogue elements of AI models can sometimes look like a marketing ploy, although Casey Newton's Platformer newsletter explains here that we need to start taking the risks a lot more seriously.

And what of other countries? Due to the massive costs involved in building LLMs, there are only two nations dominating this field – the US and China – and while others like France, UAE, Singapore, Japan and Indonesia etc might talk about developing “Sovereign AI” they only really have resources to build on top of US and Chinese systems. As the US and China are locked in a bitter tech war, with the US restricting export of its most advanced chips to China, we can't expect sensible cooperation between these two countries.

But there are huge differences in the way that the two countries are developing AI that are relevant to us all. While in the US, frontier models are proprietary systems that can only be accessed by purchasing a licence, the Chinese models (developed by companies including Deepseek, Moonshot, Bytedance, Kuaishou, Zhipu etc) are either open-source or at least open-weight, meaning that anyone can download and modify their training parameters (if not the training data) and are therefore much cheaper to use. Google, Meta and xAI have also released some open-weight models, but the flagship models in the US are largely closed.

In addition to business enterprise, US tech companies are also hoping to recoup their huge investments by ensuring that casual consumers will also be paying for AI at some point in the not too distant future (OpenAI CEO Sam Altman has already said he believes AI will become an indispensable utility that will be metered like gas, water and electricity), while Chinese companies are offering their chatbots mostly for free.

China is also paying a lot less for the electricity needed to power AI thanks to its rapidly growing use of solar and wind power generation. Not surprisingly, Chinese models are being used heavily by companies and consumers all over the world, including in the US, where they reportedly account for around one third of all AI tokens used by US firms.

At no point has that discrepancy been so clear than during the recent World Artificial Intelligence Conference (WAIC), a gathering of around 30 nations which took place July 17-20 in Shanghai. During this annual event, hosted by the Chinese government, Xi Jinping floated the idea of a World Artificial Intelligence Cooperation Organization (WAICO) to develop a multinational, open-source AI initiative that would provide an alternative to the US-led Pax Silica. He also called for AI systems to remain under human control and urged countries to establish emergency response mechanisms to manage AI risk.

So far, 29 countries have signed up for the Shanghai-headquartered organisation, most of which are located in the Global South and are not members of Pax Silica (Kazakhstan is the only country hedging its bets by signing up for both WAICO and Pax Silica). We can think of this as a digital version of China’s Belt and Road Initiative, but with LLMs and semiconductors rather than roads, power stations and ports.

Around the same time, Chinese tech company Moonshot launched its latest open-weight AI model, Kimi K3, which is said to rival frontier models like Anthropic’s Fable 5 and OpenAI’s GPT-5.6 in terms of coding, agents, architecture and size. The timing of the release, on the eve of the WAIC conference, was surely no accident.

The immediate reaction to the Kimi K3 launch was for Anthropic, OpenAI and various US government officials to collectively spit their dummies and demand restrictions on the use of Chinese open-source models, citing backdoor liabilities and other potential security issues, along with claims that China is "distilling" (basically stealing) training parameters from the US. Michael Kratsios, head of the White House Office of Science & Technology Policy, tweeted an accusation that Moonshot had covertly copied Anthropic’s Fable to produce Kimi K3.

Of course, there’s something vaguely hilarious about the fact that these US tech companies, which have basically scraped all human knowledge off the internet without asking permission, respecting copyright or paying for anything (unless dragged into hugely expensive lawsuits ), are whinging about theft.

But if Trump believes that Chinese AI is ripping off the US – and more alarmingly, could crash the stock market as investors lose confidence in the AI boom – it’s likely that restrictions could be placed on some Chinese models. That’s not something that US companies and the wider Silicon Valley community necessarily wants – it would increase their costs and there are things the US models can’t do thanks to guardrails imposed on US technology. At present it looks like two camps are forming in the US, one calling for restrictions on Chinese models, and the other arguing that excessive regulation will hinder innovation.

Chinese authorities are also weighing export controls on its own technology as a response to US regulatory pushback, with China’s Ministry of Commerce talking to local start-ups about how to prevent them being acquired by the West. And it’s worth noting that China's AI companies also haven't, as yet, developed a single, definitive and profitable business model. So there's a real chance that at some point, the protectionism might cut both ways.

In fact, we should probably take comfort from the growing realisation that the world’s biggest tech companies are throwing billions at a technology they don’t know how to monetise – because that might just put the brakes on millions of people losing their jobs or data centres contributing to the destruction of the environment – although of course we also risk our savings and pensions going up in smoke.

There are many ways this could play out – but we will hopefully get a clearer picture when the US and China hold their first AI talks ahead of Xi’s official visit to Washington in late September. In the meantime, Streamlined will be keeping a close eye on developments and advises reading Liu Cixin’s The Three-Body Problem trilogy, if you haven’t already (forget the Netflix adaptation), for an entertaining explanation of what happens when two major powers try to restrict each other’s technology. At least it will keep you occupied while the world burns.

**Xi Jinping addresses the World Artificial Intelligence Conference**

India Has Biggest First Half Year Box Office Since The Pandemic

Some good news to finish on. Ormax Media recently reported that India’s box office for the six months January to June 2026 reached $661.66m (INR63.98bn), making it the highest-grossing first half of any year since the pandemic, and surpassing the first half of 2025 by around 10%.

Even more encouraging, admissions grew by 5% during this period to 378 million, compared to the first half of 2025, reversing a three year trend of decline and stagnation. However, admissions were still below the 400 million mark achieved in the first half of 2022, a period that saw the release of two major South Indian hits: Telugu-language RRR and Kannada-language K.G.F: Chapter 2.

This half year, it’s the turn of Hindi films to lead the box office, with Dhurandhar: The Revenge topping the chart with $132m (INR12.8bn), followed by Border 2 with $39m (INR3.8bn). Telugu-language sports action drama Peddi, starring RRR’s Ram Charan, came in third with $29m (INR2.8bn), followed by another Telugu film, action comedy Mana Shankara Vara Prasad Garu, with $26.4m (INR2.6bn) and Tamil-language Karuppu with $23.7m (INR2.29bn).

While the figures are encouraging, the box office isn’t looking any less polarised with Dhurandhar: The Revenge contributing nearly 20% of total box office and the top 15 films accounting for 58% of overall receipts.

However, with several big films lined up for the second half of the year, Ormax is predicting that India box office has a chance to cross the $1.56bn (INR150bn) threshold, which will make it the best year ever, overtaking the current record set last year with $1.4bn (INR134bn). Over the last three years, the January-June period has contributed about 42% of annual India box office collections.

Films set for release in the second half include Geetu Mohandas’ Toxic, starring Yash, scheduled for August 26; Ramayana: Part 1, which is expected for Diwali in November and was recently picked up by Sony for worldwide distribution; Shah Rukh Khan starrer King, set for December 24; and Hollywod titles Avengers: Doomsday and Spider-Man: Brand New Day. Christopher Nolan’s The Odyssey opened on July 17 and has been declared a major hit.

However, Indian producers have a tendency to push release dates back multiple times, so we never really know which films are coming until they hit theatres. Sanjay Leela Bhansali’s Love And War has already been pushed into early 2027 and Salman Khan starrer Maatrubhumi: May War Rest In Peace (previously known as Battle Of Galwan) has been postponed from its original August slot with no new date given.

**Peddi, starring Ram Charan, was the highest-grossing Telugu film in 2026 H1**

IN THE TRADES:

Why China’s AI Strategy Has The US Spooked; India 2026 H1 Box Office

Liz Shackleton

US & China AI Battle: Who's Stealing From Who?

India Has Biggest First Half Year Box Office Since The Pandemic

IN THE TRADES:

LAB & FUNDING NEWS:

PRODUCTION NEWS:

CORPORATE:

CURATED:

RELEASED:

SOLD:

STREAMING UPDATES:

Read more

Purin Film Fund Awards $155k To Five Projects In Autumn 2026 Funding Round

‘Safe Exit’ & 'Citizen Osama' Take Top Awards In El Gouna

Streamlined Guides: K-Content – Korea Rents The Pipeline

BUSAN 2026: Asian Era Of Cooperation; AI Backtrack; Award Winners