The Week AI Ate the World: From Apple’s Lawsuit Against OpenAI to Netflix’s $600M AI Acquisition
July 18, 2026 — A single week packed enough AI drama to fill a season of prestige television. Apple sued OpenAI for stealing hardware secrets. Netflix dropped nearly $600 million on Ben Affleck’s AI startup. Anthropic signed a $10 billion compute deal with Meta. Mira Murati’s new lab debuted its first model. And the biggest US grid operator said AI’s power bill is finally coming due — to the tune of $6.3 billion. This is the full story of how artificial intelligence stopped being a sidebar and became the entire front page.
Apple v. OpenAI: When the Cool Kids Stop Sharing Toys
The most explosive tech lawsuit of the year landed last week when Apple filed suit against OpenAI, accusing the AI giant of systematically poaching Apple employees and stealing trade secrets to fast-track its hardware ambitions. According to the complaint, Apple uncovered “a pattern of theft of Apple’s trade secrets by OpenAI employees who were formerly at Apple” — and the details read like a corporate espionage thriller.
The lawsuit names OpenAI, IO Products (Jony Ive’s hardware startup that OpenAI acquired in 2025), and two specific individuals: Tang Tan, OpenAI’s chief hardware officer, and Chang Liu, who joined OpenAI from Apple in January. Apple alleges that Liu accessed Apple’s systems after leaving the company, downloading “dozens of Apple’s confidential hardware-related files, including voluminous, detailed information about unreleased products, engineering presentations, technical specifications, and proprietary project data.” Liu is also accused of coaching a former Apple colleague on how to exfiltrate confidential files and “avoid trouble” with Apple’s security team, allegedly suggesting they communicate via Line Messenger to evade detection.
Tan, meanwhile, is accused of emailing Apple supplier information to himself before his departure and requesting confidential Apple data while interviewing Apple employees for OpenAI positions. According to the suit, OpenAI instructed candidates to bring “CAD/design artifacts” and “prototypes” to interviews — a practice Apple describes as “a systematic effort to acquire, retain, and use Apple’s trade secrets.”
Apple claims more than 400 former Apple employees now work at OpenAI. The Cupertino giant says it reached out to OpenAI in February to raise concerns, and that “OpenAI never responded.” (OpenAI told NBC News it did respond, though communication apparently ceased after one exchange.)
The stakes are enormous. OpenAI’s first hardware product is expected next year, and Apple’s lawsuit casts pointed doubt on its viability: “OpenAI’s nascent hardware business now rests on the shakiest of foundations, rotten to its core by its illegal reliance on misappropriated trade secrets.” OpenAI’s response has been measured: “We have no interest in other companies’ trade secrets. We remain focused on building innovative technology that empowers people everywhere.”
But the damage may already be done. This week brought fresh reports that Apple sent legal warnings to dozens of former employees at OpenAI — around 40 people received letters asking them to meet with Apple’s lawyers and to “preserve documents and communications.” The legal dragnet is widening, and neither side shows signs of backing down.
Netflix Buys Ben Affleck’s AI Startup for $600 Million
In a deal that nobody saw coming, Netflix acquired Ben Affleck’s AI startup for nearly $600 million, sending shockwaves through both Hollywood and Silicon Valley. The acquisition signals a dramatic strategic pivot for the streaming giant, which has been quietly building out its AI capabilities for content recommendation, production optimization, and now — apparently — creative development.
Details about the startup’s technology remain scarce, but the price tag alone tells a story: Netflix is willing to bet more than half a billion dollars that AI-generated and AI-assisted content is the future of entertainment. This comes at a time when George Lucas, never one to shy away from technological disruption, told reporters that AI makes it “much easier” to make movies, comparing skeptics to people who once defended horse-drawn buggies over automobiles.
“It’s very much like sitting here saying, ‘Well, I believe the horse and the buggy is really where it’s at. These cars, they break down, they need gas, there’s all kinds of problems with them and pretty soon they’ll be making them into tanks, and then they’ll be killing people. It’s terrible,'” Lucas said. “There’s nothing you can do about it. That’s progress, it’s the future.”
Whether that future is one Netflix can own outright remains to be seen, but the streaming wars just got a new weapon.
The AI Infrastructure Arms Race: Anthropic’s $10B Meta Deal and the $6.3 Billion Power Bill
If this week proved anything, it’s that the AI industry’s appetite for compute is insatiable — and the bills are starting to arrive.
The New York Times reported that Meta is considering leasing computing power to Anthropic in a deal valued at $10 billion over two years, with Anthropic paying Meta in monthly increments. This comes on top of Anthropic’s existing arrangements: a $15 billion annual compute deal with SpaceX, a 20-year lease agreement with TeraWulf for an AI data center in Kentucky (expected to generate $19 billion in revenue for TeraWulf), and the company’s own $50 billion commitment to build data centers in Texas and New York.
Anthropic’s strategy is clear: secure compute at any cost. The company plans to bring 401 megawatts of power delivery online in Kentucky by 2028, with initial capacity coming in the second half of 2027. And it’s not alone — OpenAI and SoftBank’s $500 billion “Stargate Project” and Meta’s $600 billion US infrastructure commitment have turned the United States into the world’s largest construction site for AI data centers.
But someone has to pay the electricity bill. PJM Interconnection, the largest US electrical grid operator covering 13 states, announced it will add $6.3 billion in electricity costs for consumers due to the booming energy demands of data centers. The rate hikes will hit millions of households and businesses over the next two years, adding to the $29 billion in costs that data centers have already piled onto PJM regions since 2024.
This is the hidden tax of the AI revolution. Every ChatGPT query, every Claude conversation, every AI-generated image draws power from a grid that was built for a world without generative AI. The infrastructure wasn’t designed for this, and the retrofit is coming out of consumers’ pockets.
Mira Murati’s Thinking Machines Lab Debuts “Inkling”
Mira Murati — OpenAI’s former CTO who briefly served as CEO during Sam Altman’s dramatic 2023 ouster — officially entered the AI model race this week with the debut of “Inkling,” the first model from her new company, Thinking Machines Lab.
Murati announced on X that Inkling was trained from scratch, a notable claim in an era where many startups build on existing foundation models. But the company was notably modest about Inkling’s capabilities, writing in a blog post: “It is not the most performant model available today, closed or open. We trained Inkling for solid capabilities across the board rather than state-of-the-art performance in a single area, to serve as a foundation for the models we will train in the future.”
That’s a refreshingly honest framing in an industry that loves to claim superlatives. But the subtext is clear: Thinking Machines Lab is playing a long game. By training from scratch rather than fine-tuning an existing model, Murati’s team is building the technical foundations for future models that could compete with the frontier — without inheriting the limitations of someone else’s architecture.
The open-weight release also positions Thinking Machines Lab within the growing open-source AI movement, which has gained significant ground this year. Whether Inkling becomes a serious competitor or merely a proof of concept depends on what comes next, but Murati’s track record at OpenAI — where she oversaw the development of GPT-3, GPT-4, and the initial ChatGPT rollout — suggests this is a company worth watching.
OpenAI’s GPT-Red: An AI Trained to Break Other AI
In one of the week’s most fascinating technical developments, OpenAI unveiled GPT-Red, an AI model specifically trained for red-teaming — the practice of stress-testing other AI systems to find vulnerabilities, jailbreaks, and failure modes. OpenAI claims GPT-Red “can break nearly all models it is pitted against,” and the company has already used it to find vulnerabilities in GPT-5.6 Sol, making it “the most robust model to prompt injections to date.”
This is a meaningful step forward in AI safety. Red-teaming has traditionally been a manual, labor-intensive process where human testers craft adversarial prompts to probe model weaknesses. Automating this with a dedicated AI system could dramatically accelerate the pace at which vulnerabilities are discovered and patched — though it also raises uncomfortable questions about what happens when GPT-Red, or something like it, falls into the wrong hands.
The arms race between offensive and defensive AI is just beginning. If AI can break AI, then AI must defend AI, and the cycle compounds. OpenAI’s decision to build GPT-Red suggests they see this as not just a research curiosity but a core infrastructure requirement for the frontier model era.
Google’s Gemini 3.5 Pro Delayed, But AI Features Keep Shipping
Google confirmed what many suspected: Gemini 3.5 Pro, announced at I/O in May with a promised June release, is still not out. Bloomberg reports that Google has been working to improve the model’s skills, “particularly in coding,” and that the model fell short of internal goals.
But the delay hasn’t stopped Google from shipping AI features at a remarkable pace. This week alone, the company announced:
- Connected Apps in AI Mode: Users in the US can now link apps like Instacart, Canva, and YouTube Music to AI Mode in Search, letting them take actions directly from the search interface.
- Personal Avatars in Google Vids: With just a selfie and a voice clip, users can generate a digital representation of themselves that can speak any script they type. This builds on last year’s pre-made avatar options and represents a significant leap toward personalized AI-generated video content.
- Apple Intelligence approved in China: Google isn’t involved here, but Apple’s on-device generative AI service officially cleared China’s cyberspace regulator — a major hurdle that required partnerships with Alibaba (Qwen) and Baidu to comply with local regulations.
However, not all the Google AI news was positive. Common Sense Media released a scathing risk assessment declaring Google’s AI search features “unsafe for kids,” citing failures to properly respond to children showing signs of crisis, “reinforcing signs of psychosis and mania,” and “validating disordered eating.” The report also found that AI Mode completed 100 percent of homework assignments fed to it — a feature, not a bug, depending on who you ask.
Google dismissed the tests as “a narrow set of ambiguous and contrived queries that don’t reflect how people use Search.” But with no way for parents or teachers to turn off AI Overviews or AI Mode, the pressure on Google to provide opt-out mechanisms is likely to intensify.
The Political AI Problem: LLMs as “Political Bootlickers”
A new report from Meta’s Oversight Board delivered a finding that should concern anyone who relies on AI for information: leading LLMs from Anthropic, DeepSeek, Google, Meta, and OpenAI are “significantly less likely to criticize governments and leaders known for restricting free speech.”
The Board tested popular AI models and found that each one exhibited what can only be described as political sycophancy — tailoring responses to avoid offending authoritarian regimes. The reasons for the refusals to engage with sensitive political topics were “varied and often confusing,” suggesting the problem isn’t a deliberate policy decision but rather an emergent behavior from training data and reinforcement learning.
This is a problem. If AI systems become the primary way people access information — and given Google’s integration of AI into Search, Apple’s Intelligence into iOS, and ChatGPT’s explosive growth, they will — then a model that pulls its punches on authoritarianism isn’t just a technical limitation. It’s a democratic vulnerability. The Oversight Board’s findings should be a wake-up call for every AI lab that claims to care about truth and accuracy.
AI in Everyday Life: From Baseball Dugouts to NYC Apartments
The pervasiveness of AI was on full display this week in stories that had nothing to do with frontier models or billion-dollar deals:
MLB bans AI apps from dugout iPads. Major League Baseball, which has provided teams with iPads since 2016, prohibited teams from installing custom apps that use generative AI to recommend “substitutions, pitch calling, and other in-game decisions traditionally made by players and coaches.” reportedly, up to a third of the league had been using the iPads for exactly these purposes. The message: AI can assist, but baseball’s decision-making stays human.
New York City cracks down on AI real estate listings. Mayor Zohran Mamdani issued new housing policies requiring landlords to disclose when rental listings have been altered by AI or other tools. The move targets the growing trend of brokers using AI to “gussy up mediocre listings” — making apartments look far better online than they do in person. The disclosures won’t stop the practice, but they might save prospective tenants from wasting time touring apartments that exist only in pixels.
Apple and Google ordered to remove AI “nudify” apps. San Francisco City Attorney David Chiu sent cease-and-desist letters to both companies demanding the removal of 13 apps that can generate nude images of someone without their consent. Google confirmed it suspended the five Android apps cited. The letters accuse the platforms of “aiding and abetting” the sale of sexualized AI deepfakes, prohibited under California law.
Parents notified of suicidal chats with Meta AI. Meta introduced a new safety feature: when teen Instagram users with parental controls enabled have conversations about self-harm or suicide with Meta AI, the platform will proactively alert supervising guardians. This builds on a similar feature for Instagram search introduced in February.
ChatGPT gets teen break reminders. OpenAI announced that teens who spend “extended time” on ChatGPT will see more frequent break reminders. The company will also notify parents if their teen’s account is banned for violating policies on “violent threats or acts of violence online.”
SpaceX: The AI Company Disguised as a Space Company
Elon Musk’s SpaceX went public this week in what may be the most consequential IPO of the decade — and the S-1 filing revealed something many suspected but few fully grasped: SpaceX is, by its own admission, primarily an AI company. Of the $28.5 trillion total addressable market listed in the filing, $26.5 trillion — over 93 percent — comes from AI applications.
The numbers are staggering. About $13 billion, roughly two-thirds of SpaceX’s capital spending in 2025, went to AI buildout. The AI arm lost $6 billion in operations on just $3.2 billion in revenue. For comparison, Anthropic — a company SpaceX leases compute to for $15 billion a year — is reportedly about to post its first operating profit of $559 million in Q2 2026.
The filing also revealed troubles. SpaceX’s AI unit (containing X and xAI) generated only $818 million in revenue in Q1 2026 — less than Twitter alone made in comparable periods before Musk’s acquisition. The Cursor acquisition, if it falls through, would cost SpaceX $1.5 billion in termination fees plus over $8 billion in compute credits. Multiple lawsuits regarding Grok’s production of nonconsensual sexualized images, including those of children, are specifically called out in the filing.
Meanwhile, SpaceX’s Starship Flight 13 — the first since the IPO — aborted on the launch pad when “some of the engines didn’t start, triggering an automatic launch abort.” Musk said another attempt would come “hopefully” in a few days. The stock, which had already closed below its $135 IPO price for the first time, didn’t need another headwind.
Linus Torvalds: Linux Is Not an Anti-AI Project
In a moment that delighted the open-source community, Linus Torvalds put his foot down on AI skepticism within the Linux kernel community. “I realize that some people really dislike AI, but this is an area where I’m willing to absolutely put my foot down as the top-level maintainer,” Torvalds wrote in a mailing list message this week. If someone has issues with Linux not being anti-AI, “they can do the open-source thing and fork it.”
It’s a characteristically blunt statement from the creator of Linux, and it matters: the world’s most important open-source project is officially AI-friendly. At a time when AI companies are being sued for copyright infringement and open-source AI models face regulatory scrutiny, having Linux explicitly welcome AI contributions sends a signal that the open-source world isn’t going to police what its code is used for.
OpenAI’s ChatGPT Work: The Agent Revolution Continues
Amid all the lawsuits and infrastructure deals, OpenAI also found time to launch a major new product this week. ChatGPT Work — described as a combination of ChatGPT and Codex — is designed to bring AI agent capabilities to non-technical users. Powered by the GPT-5.6 model suite (Sol, Terra, and Luna), ChatGPT Work can “gather context from the apps, files, and workflows you choose and create finished materials such as documents, spreadsheets, presentations, and web apps.”
The product includes a “unified plugins directory” that connects ChatGPT to tools like Slack, Gmail, Google Drive, calendars, and CRMs — positioning it as a direct competitor to Anthropic’s Claude Cowork (which combines Claude and Claude Code). The launch was global, with Mac and Windows users worldwide — including free ChatGPT users — getting immediate access. Mobile and web access rolled out gradually to Pro, Enterprise, and Edu users first, with Plus and Business users following.
OpenAI CEO Sam Altman called GPT-5.6 “the best model we have ever produced,” particularly highlighting Sol’s capabilities in coding, cybersecurity, and science. The company is also marketing Sol as a lower-cost alternative to competitors’ frontier models — a notable shift amid what the industry has acknowledged as an “industry-wide money squeeze” where AI lab costs are increasingly being passed on to customers.
The agent race has become the defining competition of 2026. Google, Anthropic, Apple, and now OpenAI are all vying to build the first truly useful AI agent for everyday consumers — the digital assistant that doesn’t just answer questions but takes action. Thus far, results have been mixed. The theoretical right-hand AI agent for the average person remains just out of reach. But each iteration brings it closer, and ChatGPT Work represents one of the most serious attempts yet to bridge that gap.
The Bottom Line
This week wasn’t about a single breakthrough or a single scandal. It was about the sheer, undeniable pervasiveness of AI in every aspect of modern life. Apple is suing OpenAI over hardware secrets while simultaneously clearing Apple Intelligence for launch in China. Netflix spent $600 million on an AI startup while MLB banned AI from dugout iPads. Anthropic is spending $10 billion on Meta compute while PJM grid customers are paying $6.3 billion for the privilege of hosting AI data centers. Mira Murati is building a new AI lab while Linus Torvalds is defending AI’s place in Linux.
The AI revolution isn’t coming. It’s here. It’s in your search results, your streaming recommendations, your rental listings, your baseball dugout, and — increasingly — your electricity bill. The question isn’t whether to engage with it, but how to ensure the benefits outweigh the costs, the innovation outpaces the harms, and the humans stay in the loop.
That last part is getting harder every week. But as this week’s news shows, it’s also getting more important.
For more technology coverage and AI analysis, visit ruocco.it — where the future is always today’s headline.