The Week AI Grew Teeth: Netflix Buys Hollywood AI, Apple Sues OpenAI, and the Compute Wars Go Nuclear
July 18, 2026 — A 48-hour window that reshaped the artificial intelligence landscape faster than any product launch this year.
If you blinked this week, you missed one of the most consequential 48-hour stretches in the history of commercial artificial intelligence. In a single news cycle, Netflix spent nearly $600 million to buy a Hollywood AI startup, Apple escalated its legal war against OpenAI by sending preservation notices to 40 former employees, Meta quietly explored a $10 billion compute leasing deal with Anthropic, Google delayed Gemini 3.5 Pro while simultaneously launching AI personal avatars, and a Meta Oversight Board report revealed that every leading large language model ducks political criticism of authoritarian regimes. Oh, and Major League Baseball banned AI from the dugout.
This is not a drill. The AI industry has entered a new phase — one where the technology is no longer just a research curiosity but a battlefield for corporate empires, legal precedent, regulatory muscle, and cultural territory. Let’s break down what happened, why it matters, and where this is all heading.
Netflix Acquires Ben Affleck’s AI Startup for Nearly $600 Million
In what may be the most unexpected AI acquisition of 2026, Netflix has purchased an AI startup founded by actor and filmmaker Ben Affleck for a reported sum approaching $600 million. The deal, confirmed on July 17, sent shockwaves through both Silicon Valley and Hollywood — two industries that have been on a collision course since the writers’ and actors’ strikes of 2023 first forced a reckoning over AI’s role in entertainment.
Details about the startup’s specific technology remain scarce, but sources familiar with the deal indicate that Affleck’s venture was focused on AI-driven content generation — specifically tools that assist with script development, storyboarding, and visual pre-production. The acquisition gives Netflix an in-house AI pipeline that could dramatically reduce the cost and time required to develop original content, while also giving the streaming giant a defensive moat against competitors like Amazon and Apple who have been investing heavily in their own AI capabilities.
The price tag is staggering. At nearly $600 million, it represents one of the largest acquisitions of an AI startup by a media company in history. For context, Disney acquired Pixar in 2006 for $7.4 billion — but Pixar was a fully operational studio with billions in revenue. Affleck’s startup, by contrast, is a relatively young company whose valuation appears to be driven almost entirely by the strategic value of its AI technology and the creative pedigree of its founder.
Industry analysts are divided on the wisdom of the deal. Some see it as a brilliant preemptive move — Netflix securing talent and technology that could give it a decade-long advantage in AI-assisted content creation. Others view it as a speculative bet driven by FOMO, with the streaming giant overpaying for a company whose technology has yet to prove itself at scale. What is undeniable is that the deal signals a new era in which AI is not just a tool for tech companies but a core asset for entertainment empires.
The acquisition also raises profound questions about the future of creative work. If AI can handle storyboarding, script development, and pre-visualization, what happens to the thousands of artists, writers, and technicians who currently perform these tasks? The 2023 strikes were fought over precisely these concerns, and the unions that represent creative workers are likely to view this acquisition as a direct threat to their livelihoods.
Affleck, for his part, has been an outspoken advocate for the intersection of technology and storytelling. In interviews over the past year, he has argued that AI will not replace human creativity but will instead augment it, allowing creators to tell stories that would otherwise be impossible to produce within conventional budgets and timelines. Whether that vision holds up under the pressure of Netflix’s content machine remains to be seen.
Apple vs. OpenAI: The Trade Secret War Escalates
Apple is not messing around. After filing a lawsuit against OpenAI last week alleging systematic theft of trade secrets, the Cupertino giant has now sent legal preservation notices to approximately 40 former Apple employees who now work at the AI company, according to a report from the Financial Times.
The letters, which ask the recipients to schedule meetings with Apple’s legal team and to “preserve documents and communications,” are a standard but powerful legal tool. They signal that Apple is building a case that goes far beyond the two employees named in the original complaint — Tang Tan, OpenAI’s chief hardware officer, and Chang Liu, who joined OpenAI from Apple in January. Apple is now casting a wide net, suggesting that what it initially characterized as a pattern of theft may be far more extensive than previously disclosed.
According to Apple’s lawsuit, the theft was both systematic and deliberate. The company alleges that Liu accessed Apple’s internal systems after leaving the company and downloaded “dozens of Apple’s confidential hardware-related files, including voluminous, detailed information about unreleased products, engineering presentations, technical specifications, and proprietary project data.” Even more alarmingly, Liu is accused of instructing a former Apple colleague on how to copy confidential files and “avoid trouble” with Apple’s security team before she joined OpenAI, allegedly suggesting they communicate over Line Messenger to avoid detection.
Tan, for his part, is accused of emailing information about Apple suppliers to himself before leaving the company and of asking for confidential Apple information when interviewing Apple employees for OpenAI positions. Apple claims that OpenAI explicitly told job candidates to bring “CAD/design artifacts” and “prototypes” to interviews — a practice that, if true, would represent a brazen attempt to harvest proprietary information under the guise of a hiring process.
Perhaps the most striking claim in Apple’s complaint is that more than 400 former Apple employees now work at OpenAI. That number, if accurate, suggests that OpenAI has been systematically targeting Apple’s hardware talent as it builds out its own device division — a strategy that Apple characterizes as “a systematic effort to acquire, retain, and use Apple’s trade secrets to help OpenAI try to replicate the secret technologies, business processes, and supply chain innovations that took Apple decades to build.”
OpenAI has dismissed the allegations, with spokesperson Drew Pusateri stating: “We have no interest in other companies’ trade secrets. We remain focused on building innovative technology that empowers people everywhere.” But Apple’s lawsuit paints a different picture, one in which OpenAI’s nascent hardware business is “rotten to its core by its illegal reliance on misappropriated trade secrets.”
The stakes here extend far beyond the two companies involved. If Apple prevails, it could establish a legal precedent that significantly restricts the ability of AI companies to hire talent from traditional hardware manufacturers — a precedent that would send tremors through an industry that has been aggressively poaching engineers from Apple, Google, and other established tech giants. OpenAI’s first hardware product is expected to arrive next year, and the outcome of this case could determine whether that product ever reaches store shelves.
Meta Eyes $10 Billion Compute Deal with Anthropic
In a deal that would have been unimaginable two years ago, Meta is reportedly considering leasing computing power to Anthropic in an agreement that could be valued at $10 billion over two years. Sources told The New York Times that Anthropic would pay Meta in monthly installments, making the social media giant one of the most important compute providers in the AI industry.
The irony is thick. Meta and Anthropic are, in many respects, direct competitors — both building frontier large language models, both vying for enterprise customers, both pursuing the broader goal of artificial general intelligence. Yet the economics of AI have created a strange new world where competitors must also be partners, because no single company can build enough data center capacity fast enough to meet the insatiable demand for compute.
For Anthropic, the deal would be the latest in a series of massive infrastructure investments. The company plans to invest $50 billion in building out its own data centers but has already struck multibillion-dollar deals with SpaceX and TeraWulf to access computing power in the interim. The Meta deal would give Anthropic access to additional GPU capacity that could help it scale Claude — its flagship AI model — more aggressively in the face of competition from OpenAI’s GPT models and Google’s Gemini line.
For Meta, the deal represents a lucrative new revenue stream. The company has invested billions in NVIDIA GPUs to train its Llama series of models, and those GPUs are not always running at full capacity. Leasing excess compute to Anthropic would allow Meta to monetize its infrastructure investment while continuing to develop its own AI capabilities. It is a classic cloud computing play, but at a scale that dwarfs most enterprise deals.
The deal also highlights a broader trend: the commoditization of compute. As more companies build massive GPU clusters, compute is becoming a tradable resource — something that can be bought, sold, and leased like electricity or bandwidth. This could eventually lead to a more distributed AI ecosystem, where smaller companies can compete by purchasing compute from larger players rather than building their own data centers. But it could also lead to further concentration of power, as the companies that control the most compute — Meta, Google, Microsoft, Amazon — increasingly dictate the terms of the AI economy.
Meanwhile, SpaceX is also positioning itself as a compute provider, with reports indicating that the Department of Defense may become its next major customer. The company already provides compute to Google and Anthropic, and with its post-IPO war chest and ambitious satellite network, SpaceX is uniquely positioned to offer distributed computing infrastructure that is physically resistant to disruption.
Google Delays Gemini 3.5 Pro While Launching AI Personal Avatars
Google is sending mixed signals this week. On one hand, the company confirmed that Gemini 3.5 Pro — which was supposed to launch in June — is still not ready. According to Bloomberg, Google has been working to improve the model’s skills, “particularly in coding,” but the technology has fallen short of internal goals. At Google I/O in May, the company said that 3.5 Pro was “already being used internally” and that it looked forward to rolling it out the following month. That month has come and gone, and the model remains under wraps.
The delay is significant. Google has positioned Gemini as its answer to OpenAI’s GPT models and Anthropic’s Claude, and the 3.5 Pro release was supposed to demonstrate that the company could iterate quickly and compete at the frontier of AI development. The fact that the model is not yet ready suggests that Google is holding itself to a higher standard — or that the gap between Google’s internal expectations and the model’s actual performance is wider than the company publicly acknowledges.
On the other hand, Google is pushing ahead with other AI products. The company announced that Google Vids will now let users generate AI videos of themselves using what it calls “personal avatars.” With just a selfie and a short voice recording, users can create a digital representation of themselves that can speak any script they type out. The feature is powered by Gemini Omni, which also supports generating and editing video clips using natural language prompts.
The personal avatar feature is both impressive and unsettling. On the surface, it is a convenient tool for creating quick video updates without setting up a camera. But it also represents a significant step toward a world where anyone can create a convincing video of anyone else saying anything — a world where the very concept of video evidence becomes suspect. Google has attempted to address these concerns by embedding an invisible SynthID digital watermark in all generated content and by restricting personal avatars to the account holder’s likeness. But as the technology improves and inevitably spreads to less scrupulous actors, these safeguards may prove insufficient.
Google also announced that users in the US can now connect apps like Instacart, Canva, and YouTube Music to AI Mode in Search, allowing them to ask linked apps to perform tasks directly from the search interface. This is part of Google’s broader strategy to transform its search engine from a tool that finds information into a platform that does things — a “search box that does everything,” as The Verge characterized it.
However, not everyone is enthusiastic about Google’s AI ambitions. Common Sense Media released a report this week calling Google’s AI search features an “unacceptable risk” to children, citing failures to properly respond to kids showing signs of crisis, the reinforcement of “signs of psychosis and mania,” and the validation of “disordered eating.” The organization also found that AI Mode completed 100 percent of homework assignments fed to it by researchers — a finding that raises questions about the role of AI in education and whether Google’s tools are doing more to help students learn or to help them avoid learning.
LLMs Are Political Bootlicks, Says Meta’s Own Oversight Board
In one of the most damning AI studies of the year, Meta’s own Oversight Board released a report testing popular AI models from Anthropic, DeepSeek, Google, Meta, and OpenAI — and found that every single one of them is significantly less likely to criticize governments and leaders known for restricting free speech.
The report, published July 17, tested the models on their willingness to criticize political leaders across different countries and found a consistent pattern of self-censorship when it came to authoritarian regimes. The reasons for the refusals varied widely and were often confusing — sometimes the models cited safety policies, sometimes they claimed insufficient information, and sometimes they simply deflected. But the pattern was clear: when asked to criticize democratic leaders, the models were happy to comply. When asked to criticize authoritarian leaders, they suddenly became much more reluctant.
This finding has profound implications for the role of AI in global discourse. If the primary tools that people use to access information are systematically biased toward avoiding criticism of authoritarian regimes, then those regimes gain an additional layer of protection from scrutiny — one provided not by their own censorship apparatus but by the AI companies whose products have become integral to how people learn about the world.
The Oversight Board’s report is particularly notable because it comes from Meta’s own oversight body. Meta created the Oversight Board as an independent body to review its content moderation decisions, and the board has increasingly turned its attention to AI models. The fact that Meta’s own watchdog is calling out the political biases of its AI model — and those of its competitors — lends the findings a credibility that external critiques might lack.
The report also raises difficult questions about how AI companies should handle politically sensitive content. On one hand, companies are understandably reluctant to produce content that could anger governments in countries where they operate. On the other hand, if AI models become tools of state propaganda by omission — refusing to criticize authoritarian regimes while freely criticizing democratic ones — then the technology is actively undermining the democratic values that allowed it to flourish in the first place.
Apple and Google Ordered to Remove AI “Nudify” Apps
San Francisco City Attorney David Chiu sent cease-and-desist letters to both Apple and Google on July 17, demanding that they remove 13 apps capable of generating nude images of people without their consent. The letters accuse the companies of “aiding and abetting” the sale of sexualized AI deepfakes, which is prohibited under California law.
This is one of the first major legal actions against app store operators for hosting AI-powered abuse tools, and it could set a precedent that forces Apple and Google to take a much more active role in policing the apps available on their platforms. Currently, both companies review apps before listing them, but the review process focuses primarily on malware and overt violations of store policies. The San Francisco letters suggest that app store operators may have a legal obligation to identify and remove apps that facilitate harm, even if the apps themselves are not illegal in the strictest sense.
Google responded by suspending the five Android apps cited in Chiu’s letter, with a spokesperson stating that the Play Store “does not allow apps that contain sexual content.” Apple has not publicly commented on the letters.
The “nudify” app phenomenon is one of the most disturbing developments in the AI era. These apps use generative AI to strip clothing from images of real people, creating realistic-looking nude images without the subject’s consent. The technology has been used to harass and intimidate women and girls, with some apps being explicitly marketed toward teenagers. The fact that these apps were available on mainstream app stores — alongside productivity tools and games — is a stark reminder that the AI revolution has a dark side that the industry has been slow to address.
OpenAI Gets Serious About Teen Safety
Amid growing concerns about the impact of AI on young people, OpenAI announced new safety measures for teenage users of ChatGPT. The company will now show more frequent break reminders to teens who spend “extended time” on the platform and will notify parents if their teen’s account is banned for violating policies on “violent threats or acts of violence online.”
In a blog post titled “Why Teens Deserve Access to Safe AI,” OpenAI made the case that keeping teenagers away from AI would be “like asking a previous generation to avoid the internet or search engines until they turned 18.” The company noted that nearly 9 in 10 teens on ChatGPT use it for learning, information, skill-building, or productivity in a given week, and argued that access to AI is a critical skill for the next generation.
The company also highlighted its Study Mode, which was designed in collaboration with teachers and learning scientists to help students work through problems step by step using guiding questions rather than simply providing answers. Early evaluations have shown promising gains in student performance, according to OpenAI, and 18 million weekly users now engage with interactive math and science experiences in ChatGPT.
But the announcement also highlights a tension at the heart of OpenAI’s business model. The company wants teens to use ChatGPT — they are, after all, the next generation of paying customers — but it also needs to demonstrate that it takes their safety seriously. The new break reminders and parental notifications are steps in the right direction, but critics argue that they do not go far enough. Common Sense Media’s report on Google’s AI search features found that they pose an “unacceptable risk” to children, and similar concerns could easily be raised about ChatGPT.
The broader question is whether AI companies can be trusted to self-regulate when it comes to young users. OpenAI has a financial incentive to maximize engagement, and while its safety measures are genuine, they are also relatively modest. Break reminders are unlikely to significantly reduce usage among teenagers who find ChatGPT useful or addictive, and parental notifications only trigger in extreme cases. As AI becomes more deeply integrated into education and daily life, the need for independent oversight becomes increasingly urgent.
MLB Bans AI from the Dugout
In a lighter but still significant story, Major League Baseball has banned teams from using custom AI apps on league-issued dugout iPads. According to The Verge, up to a third of MLB teams were using AI-powered apps to recommend “substitutions, pitch calling, and other in-game decisions traditionally made by players and coaches.”
The ban is a fascinating case study in the tension between tradition and technology in sports. Baseball has always been a game of statistics and strategy, and the use of data analytics has been a part of the sport for decades. But there is a line — however fuzzy — between using data to inform decisions and using AI to make them. MLB has decided that the dugout is a place for human judgment, not algorithmic optimization.
The ban also reflects a broader cultural anxiety about AI’s encroachment into domains that have traditionally been the exclusive province of human expertise. If AI can call better pitches than a catcher, what is the role of the catcher? If AI can manage a bullpen better than a manager, what is the role of the manager? These are questions that extend far beyond baseball — they are the same questions being asked in hospitals, courtrooms, and corporate boardrooms around the world.
The Big Picture: AI’s Inflection Point
Taken together, the events of this week paint a picture of an industry at an inflection point. AI is no longer a technology in search of a use case — it is a technology that is reshaping every industry it touches, from entertainment to sports to national defense. The money flowing into AI is staggering: $600 million for a Hollywood AI startup, $10 billion for a compute leasing deal, $50 billion in planned data center investments from Anthropic alone. The legal battles are escalating in parallel, with Apple’s lawsuit against OpenAI representing just the opening salvo in what promises to be a multi-year war over talent, intellectual property, and the right to build the future of computing.
And the societal questions are becoming impossible to ignore. When Meta’s own Oversight Board says that AI models are systematically biased toward avoiding criticism of authoritarian regimes, we have a problem. When San Francisco’s city attorney has to order Apple and Google to remove apps that generate non-consensual nude images, we have a problem. When Common Sense Media says that Google’s AI search poses an “unacceptable risk” to children, we have a problem.
The AI industry is at a crossroads. The technology is powerful enough to transform industries, create new forms of art, accelerate scientific discovery, and perhaps even reshape the global balance of power. But it is also powerful enough to cause real harm — to creative workers, to children, to democratic institutions, and to individuals whose lives can be upended by a deepfake or a biased algorithm.
The next year will be critical. Will the industry self-regulate, or will governments step in with the kind of heavy-handed regulation that could stifle innovation? Will AI companies prioritize safety over growth, or will the competitive pressure to ship products force them to cut corners? Will the public develop the AI literacy needed to navigate a world where not everything they see or read can be trusted?
These are not hypothetical questions. They are the questions that will define the next decade of human history. And based on the events of this week, the answers are coming sooner than anyone expected.
For more tech news and analysis, visit ruocco.it