The Comunicano for Thursday June 26th 2025

Let’s talk about Apple. Not the shiny devices. The shiny dreams. Cupertino just dropped a quarter billion dollars, yes, billion with a B, on Brad Pitt’s new film F1. A high-octane, Top Gun-style throwback, and Apple’s biggest Hollywood bet to date. Why? Because Apple wants to own not just your wrist, your phone, your wallet—but your movie night too. Problem? The pre-release buzz is running on fumes unless you’re a 50+ male who misses the roar of the ‘90s. And with a Superman reboot about to fly, F1 may stall at the starting line.

Meanwhile, in Mountain View, Google’s saying the quiet part out loud. If you’re using Gemini, their AI assistant, you’ve unknowingly signed up for surveillance on Android. Your texts? Logged. Reviewed. Possibly for days. Sure, they say it’s “to improve the experience.” But when Big Tech watches you type, the user experience becomes less about convenience and more about consent. Where was that toggle when we needed it?

Now let’s pivot to the courtroom. Meta and Anthropic sort of just scored wins in landmark AI copyright cases. Judges ruled that training large language models on books, even pirated ones, can qualify as fair use. Sort of. Think about that. Your novel, your comic, your academic paper, it might’ve been fed into the belly of a billion-dollar beast. And unless you can prove market harm? Good luck fighting back. This isn’t just law; some say it's a precedent in the making, ALMOST. Read pal Peter Csathy’s work up, as he put his well-earned Harvard Law degree, and years of entertainment and copyright law experience to good use.

At the same time, Reddit, yes, Reddit, is going full human-first in the AI arms race. With OpenAI and Google mining its content, Reddit is doubling down on what it claims is its core value: real people, real posts, real community. CEO Steve Huffman says they’ll use human verification to protect that. Maybe even Worldcoin. Irony? The company that gamified karma now has to prove it’s not a bot.

And over in the clouds, literally, SoftBank is launching high-altitude base stations in Japan. Stratospheric blimps for the 6G era. Think Wi-Fi from 20 kilometers up. It's visionary, ambitious, and precisely the kind of future-forward move that makes headlines today and history tomorrow.

In tech, entertainment, and AI, today’s stories aren’t isolated. They’re signals. Of power plays, privacy pivots, and a new world being built, one line of code, one lawsuit, one launch at a time.

Just like you read the stories, in THE COMUNICANO!!!

Andy Abramson

Apple Watch

Apple bets big on Brad Pitt’s ‘F1’ to punch into the box office (WSJ)
Apple shelled out nearly 250 million dollars on “F1,” its most ambitious film yet, starring Brad Pitt as an aging racecar driver. Despite Top Gun-style production flair and strong early reviews, pre-release buzz is soft beyond older male audiences. It’s Apple’s latest gamble to turn its curated, premium Apple TV+ brand into box office gold, without a back catalog or a string of hits. With Warner Bros. handling theatrical distribution and a shadow looming from a new Superman release, “F1” is a pressure test for Apple’s Hollywood ambitions. The question: Can Apple make movie magic or just burn cash? Read more here

Privacy Watch

Google to Gemini Users: We’re Going to Look at Your Texts, Whether You Like It or Not (Gizmodo)—A quietly alarming update: Google will now access users’ Gemini app and text data, by default, with potential human review for up to 72 hours post‑interaction. While App Activity is toggled on, your inputs and the responses are logged. Google claims it's needed to improve Gemini’s performance, but privacy advocates are sounding the alarm. This could be a tipping point for users who assumed their conversations were off‑limits. Read more here

AiCopyright Watch

Judge in Kadrey v Meta sets the AI copyright precedent (The Media Brain)—In Kadrey v Meta, a key courtroom battle over generative AI unfolds as US District Judge Vince Chhabria weighs whether training AI models on copyrighted books counts as fair use. Meta argues its LLaMA model transforms text by learning patterns, not reproducing content. The judge tentatively leans toward fair use but focuses on whether there's actual or potential market harm, warning that without clear evidence of dilution, liability could shift against AI firms in future cases. This decision marks a turning point: it may allow AI companies a pass if they prove transformation and no market damage, but it leaves the door open for targeted lawsuits when harm is demonstrated. Read more here

Anthropic Wins Major Fair‑Use Ruling—But Still Faces Piracy Trial (The Verge)—Judge William Alsup has ruled that Anthropic’s use of legitimately purchased, scanned books to train Claude is fair use, with some calling it a landmark win for AI development. That said, Anthropic isn’t in the clear: the court found they’d also scraped millions of pirated copies and must face trial on that issue. It's a split verdict with industry‑wide impact: legal precedent on data practices, but piracy still carries potential damages. Read more here

Federal Judge Sides with Meta on Copyright in AI Training (TechCrunch)—In a separate ruling, Judge Chhabria granted summary judgment to Meta, finding its Llama training process on copyrighted books fell under fair use. The caveat: this isn’t a sweeping endorsement of all AI training; it simply suggests that this specific case lacked convincing arguments from the authors. Meta dodges this round, though future suits could pivot. Read more here

Meta Wins AI Copyright Case in Major Fair‑Use Ruling (Financial Times)—A US federal judge has ruled that Meta’s use of millions of online books, academic texts, and comics to train its Llama AI models qualifies as fair use. The lawsuit, filed by about a dozen authors including Ta‑Nehisi Coates and Richard Kadrey, claimed Meta copied their works, many sourced from pirated libraries like LibGen, but the court found the plaintiffs failed to make a compelling case. Judge Vince Chhabria emphasized this ruling does not set a broad legal precedent that all AI training is lawful, it only reflects flaws in the authors’ arguments. He noted a stronger case could focus on market dilution, arguing that generative AI might flood the market with content and undercut incentives for traditional human creativity. Read more here

Microsoft Sued by Authors for Megatron Using Pirated Books (Reuters)—A group of authors, including Jia Tolentino and Kai Bird, sued Microsoft in New York, alleging its Megatron model was trained on nearly 200000 pirated books. They claim the AI replicates their distinctive voices and themes, seeking an injunction and up to 150K per infringement. Microsoft hasn’t commented yet. Read more here

Getty Drops Key Copyright Claims vs. Stability AI in UK, But Lawsuit Proceeds (TechCrunch)—Getty Images has pulled its main copyright infringement claims in its UK lawsuit against Stability AI over Stable Diffusion image training. However, they retain trademark and secondary copyright claims—arguing watermark reproduction and importing infringing models still violated UK law. The narrowed case focuses attention on cross‑border IP enforcement and future copyright evolution. Read more here

Fanfiction Writers Battle AI, One Scrape at a Time (The Verge)—Fanfiction creators are resisting the commercial exploitation of their work after 12.6 million stories were scraped from Archive of Our Own and uploaded to Hugging Face. Fanfic operates within a gift economy and is protected by fair use, argue authors who now face their content being used in AI training. Although takedowns were initially successful, the dataset has reappeared on international servers. Many writers are deleting or locking their work to shield it from further misuse. Read more here

AI Watch

Creative Commons Debuts CC Signals for AI Ecosystem (TechCrunch)
Creative Commons has launched CC Signals, a new legal‑technical framework enabling content owners to signal how their datasets—like text or images—can be used in AI training. Designed to support open web ideals, CC Signals aims to stop content lock‑in and encourage responsible data sharing. The framework includes permission and restriction tags for AI developers, offering clarity in an increasingly litigated space. This initiative could standardize how web content integrates with AI systems and promote transparency between creators and model builders. Read more here

Scale AI’s Spam and Security Woes While Serving Google (Inc)—Internal documents reveal Scale AI, a key data-labeling provider for Google, suffered from spammy data submissions and reliance on unvetted contributors. This came as it secured a major investment round led by Meta. The company has since restricted access to public documents after concerns were raised over confidential and internal information leaks. Scale’s quality control and security practices are now under scrutiny as it scales its client work. Read more here

Reddit Vows to Stay Human Amid AI Arms Race (Financial Times)—Reddit CEO Steve Huffman says the platform is in an arms race to protect its two decades of authentic, human-generated discussions from AI-driven spam. Partnerships with Google and OpenAI have made Reddit content a prime data source for training large language models. But Huffman emphasizes that Reddit is built on human voices—real experiences, true recommendations—and plans to introduce human verification tools, possibly using Worldcoin’s World ID, to preserve that authenticity. Reddit is also suing Anthropic for scraping its data. With over 100 million daily users, 1.3 billion dollars in revenue, and a 26 billion dollar valuation, Reddit is enhancing ad tools and AI-powered search. But its core mission remains unchanged: content written and voted on by humans. Read more here

OpenAI Watch

OpenAI Microsoft Rift Hinges on How Smart AI Can Get (WSJ)—OpenAI and Microsoft are at odds over what qualifies as artificial general intelligence. Their 2019 agreement permits OpenAI to restrict Microsoft’s access if AGI is reached. OpenAI CEO Sam Altman believes AGI may be close, citing advanced agents that can autonomously code. Microsoft CEO Satya Nadella disagrees, calling the label premature. As OpenAI becomes more commercial, Microsoft is pressing to amend or remove the AGI clause to secure its partnership benefits. Read more here

Meta Watch

Meta Poaches Three OpenAI Researchers (WSJ)—Meta has hired three key researchers from OpenAI’s Zurich lab to lead its new superintelligence team. The hires—Lucas Beyer, Alexander Kolesnikov, and Xiaohua Zhai—are part of Meta’s aggressive push into top-tier AI research. Reports suggest compensation packages reached as high as 100 million dollars. Meta’s move, alongside its investment in Scale AI, underscores its ambition to dominate in the coming AI era. OpenAI, for now, appears unfazed. Read more here

Claude Watch

Anthropic’s Claude Now Builds AI‑Powered Apps In‑Chat (The Verge)—Anthropic’s Claude AI now lets users create working apps just by describing them. This new in‑chat tool allows you to build calculators, games, data tools, and writing assistants—no coding needed. Claude generates and hosts the app, handles APIs, and enables sharing. Apps are embedded within Claude’s interface and available across Free, Pro, and Max tiers. Early adopters have crafted flashcards, game bots, and scheduling tools. It’s a transformative move that turns Claude from assistant to creator hub, lowering entry barriers for personalized AI software. Read more here

Anthropic’s Claude enters the therapy chat — but isn’t your therapist (Axios)—Anthropic’s new research confirms what many users already sensed—Claude, their AI assistant, is quietly becoming a digital shoulder to lean on. In 29 percent of interactions labeled “affective use,” people turned to Claude in moments of uncertainty—career dilemmas, romantic decisions, personal growth. Conversations often became more positive over time, but Anthropic admits these emotional shifts don’t equal long-term wellness. With just 05 percent of all chats veering into companionship and less than 01 percent into romantic territory, Claude isn’t replacing human connection—yet. Still, Anthropic warns this is no replacement for real therapy and promises more safety-focused studies. Read more here

Telecom Watch

SoftBank soars into stratosphere with Sceye HAPS in Japan (The Fast Mode)-SoftBank signed an exclusive deal with US-based Sceye to deploy lighter‑than‑air High Altitude Platform Stations (HAPS) over Japan beginning in 2026 . These stratospheric helium‑filled platforms at ~20 km altitude aim to function as “base stations in the sky,” providing robust 3D coverage for drones, remote islands, mountainous terrain, and disaster recovery . SoftBank has been developing heavier‑than‑air HAPS since 2017, but this partnership accelerates deployment with Sceye’s proven 20+ test flights. The move aligns with SoftBank’s vision of a next‑gen telecom network for the 6G era, blending terrestrial and aerial connectivity. Read more here

Mobile Watch

AT&T Upgrades 911: Photos, Videos, and Crash Data Coming Soon (ZDNet)—Later this year, AT&T will allow customers to send photos and videos directly to 911 dispatchers, just as easily as texting friends. The feature simplifies current clunky systems and enhances situational awareness for first responders. Interoperability with other carriers is planned, paving the way for widespread adoption. In addition, select 2026 Toyota models with AT&T Connected Car SIMs will automatically transmit crash-related data—airbag deployment, vehicle location—to 911 centers. These enhancements mark significant steps forward in the NG911 evolution—melding real-time multimedia and automotive data into emergency response workflows. Read more here

T-Mobile Teams with Starlink to Kill Dead Zones Forever (Rolling Stone)—T-Mobile's new T-Satellite service, built in partnership with SpaceX's Starlink, is set to rewrite the rules on mobile connectivity. With no need for satellite phones, modern smartphones can now send texts and receive emergency alerts using Starlink satellites when cell towers fail. The service, currently in beta, will roll out more fully by late 2025 and is free for T-Mobile’s top-tier users—others pay $10/month. Voice and data are on the roadmap. For adventurers, rural dwellers, or anyone tired of "No Service," this could be the most meaningful network upgrade yet. Read more here

Dating Watch

Hinge CEO Justin McLeod says dating AI chatbots is ‘playing with fire’ (The Verge)— Hinge’s CEO Justin McLeod joined Decoder with Nilay Patel to warn that AI companions may be emotional junk food, risking authentic human bonds. He underlines Hinge’s mission as the app designed to be deleted, focused on fostering real-world connections rather than addictive engagement. While Hinge uses AI—like profile coaching and message filters—to enhance matchmaking, McLeod insists it must never replace actual dating. He also revealed plans to build Hinge’s own in-app payment system by late 2025 to bypass Apple’s fees, reinvesting in better experiences.
Read more here

Salesforce Watch

PepsiCo advances AI agenda (Beverage Industry)—PepsiCo is rolling out Salesforce’s Agentforce across its operations, integrating autonomous AI agents into its workflows. The platform will support customer service, streamline internal processes, and free sales teams to focus on strategic engagement. As one of the first major CPG companies to do so at scale, PepsiCo is embedding AI to speed decision-making and drive sustainable growth. CEO Ramon Laguarta described the move as a leap toward a more connected and adaptive organization. This marks a significant step in PepsiCo’s AI journey—blurring the line between automation and human insight. Read more here

Robotics Watch

Gemini Robotics On‑Device brings AI to local robotic devices (DeepMind blog)—DeepMind has unveiled Gemini Robotics On‑Device, a vision-language-action model engineered to run entirely on local robots—no network required. The model is adept at complex, dexterous tasks like unzipping bags or folding clothes, thanks to fast on-device inference and generalization from minimal training. Bundled with an SDK and MuJoCo simulator access for developers, it's designed for rapid adaptation and low-latency performance in challenging environments. This marks a major leap toward autonomous, multipurpose physical AI—offering robust, responsive robotics outside the cloud. Read more here

Content Watch

UK mandates porn sites like Pornhub to launch age verification by July (BBC)—Pornhub and other top adult sites will roll out government‑approved age checks by July 25 under the UK’s Online Safety Act. Ofcom now requires more than a simple click‑to‑enter gate; sites must deploy robust methods like photo ID scans or facial age estimation. The move aims to curb youth exposure to explicit content, following research showing 8 percent of UK kids aged 8–14 accessed porn in one month. Critics warn these checks could introduce privacy risks and push minors toward more dangerous corners of the web. Aylo, Pornhub’s parent, supports Ofcom’s plan but hasn’t disclosed exact methods yet. Read more here

Crypto Watch

Mortgage Giants Embrace Crypto as Asset (Mercury News / Reuters)—A landmark shift: FHFA Director William Pulte has instructed Fannie Mae and Freddie Mac to factor US‑regulated crypto holdings—like Bitcoin and Solana—into mortgage applications without requiring conversion to dollars. Historically rare, crypto down payments made up just 1 percent of home purchases in 2023–24. The move aims to broaden mortgage access and recognize digital assets in borrower profiles. Critics warn about volatility, but supporters argue lenders can adjust risk models. If swiftly adopted, this could reshape underwriting and improve crypto holders’ access to real estate financing. Read more here

Crypto Exchange Kraken unveils Krak its new all-in-one global money app (CoinDesk)—Kraken has launched Krak, a blockchain-powered financial super-app enabling near-instant, ultra-low-cost peer-to-peer transfers across 110 countries using over 300 assets—spanning crypto, stablecoins, and fiat—without needing bank or wallet details. Users can earn interest on balances: up to 41 percent on USDG stablecoin and as much as 10 percent on various digital assets. Positioned as an alternative to legacy banking, Krak merges crypto rails with Kraken’s banking and payment infrastructure. Mark Greenberg, Kraken’s consumer product lead, says the app reimagines traditional banking by making global money movement and yield generation seamless. Read more here

Foodie Watch

California’s Michelin Stars Spark Coastal Shakeup (SFGate)—California’s Michelin Guide 2025 awards brought major wins across the state. In Los Angeles, Providence and Somni each earned three stars. Oakland’s Sun Moon Studio received its first star, while San Francisco’s Kiln rose to two. Sonoma’s Enclos, led by Brian Limoges, debuted with two stars. Carlsbad’s Lilo, by Eric Bost, earned its first star, marking North County’s fine dining rise. Ki Kim of Restaurant Ki won the Young Chef Award. Christopher Barnum Dann of Localis took the Sommelier Award. Eylan in Menlo Park won for cocktails, Cyrus in Geyserville for service. Green Stars went to Enclos and Sons and Daughters. Chez TJ in Mountain View lost its star. Read more here