📰 Welcome back!

Greetings to another mad week in the world of AI, and by the looks of it, everything else.

At least we’re another ray of sunlight in it, and as ever, more than pleased to bring you a round-up of the week’s biggest AI stories, including OpenAI’s $300 billion Stargate deal with Oracle, Musk’s conspiracy rabbit hole(s), and Google’s DOGE spaz out.

Let’s get down to business.

🚀 What we’re covering today…

• 🔫 Musk says OpenAI whistleblower was murdered

• 🕵️‍♀️ Charlie Kirk’s killer hunted with AI

• 👩🏻‍🍳 Meet R1, the new Iron Chef

• ✨ OpenAI signs $300 billion Stargate contract

• 📜 Ted Cruz introduces ‘Sandbox’ legislation

• 🔎 Google denies DOGE exists

• 🏴‍☠️ Anthropic anti-piracy lawsuit stalls

• 🍎 Apple AirPods are lost in translation

• 📊 Bankman-Fried added to new Hall of Shame

🔴 Quick Note: We like to cover loads of AI news in our newsletter, so for a better reading experience, we suggest opening this in your browser for the full experience! 

Head to the ‘READ ONLINE’ tab at the top of this email.

👁️ 👁️ What you might have missed

• After a global dumpster fire of a week, we thought it would be fitting to start this one with the new nadir in which the Musk-Altman beef has reached. “He was murdered,” Elon wrote on X, referring to the death of OpenAI whistleblower Suchir Balaji, who was ruled to have committed suicide last year via a self-inflicted gunshot wound. Balaji, a 26-year-old software engineer at OpenAI had allegedly told the New York Times he was willing to testify about illegal copyright practices at the Altman-led tech firm, with Musk’s implication being he was murdered to prevent him doing so. Musk’s comment follows Altman tip-toeing around the subject on Tucker Carlson’s show, and have only gone fuel conspiracy theories about Balaji’s death.

• In actual assassination news, digital detectives have called on AI to do some heavy lifting in the manhunt for the killer of activist Charlie Kirk. Numerous people on social media immediately used AI tools to “upscale” or “enhance” two images of the suspected killer, Tyler Robinson, which were released by the FBI last week. Some used X’s Grok bot and others used tools like ChatGPT, producing sharper, higher-resolution versions, but while these AI-enhanced versions may look more detailed, they often introduce artifacts and fabrications – imaginary features that weren’t in the original. Experts warn that these AI creations aren’t reliable as evidence in an investigation, and may do more to hinder it overall.
  

• Ant Group, the company behind Alipay and owned by Jack Ma, has unveiled its first humanoid robot, R1, via its robotics arm, Robbyant. The Chinese robotics company is positioning itself as a challenger to Tesla’s Optimus, and R1 made a pretty compelling case after its debut at IFA 2025 in Berlin. The bot went all Benihana as it prepared shrimp in front of an audience – not quite a full Iron Chef, but Ant envisions R1 doing roles like caregiver companion, travel guide, or serving in restaurants are all possible. No launch date or pricing has been revealed, with observers cautioning it’s still in a slow and preliminary phase, much like I am in the kitchen.

• Stargate SG1, HWFG: OpenAI has entered into a five-year contract with Oracle under the code-name Project Stargate, beginning in 2027, to purchase about $300 billion worth of cloud computing capacity – one of the largest in history. In July, the partnership was unveiled alongside SoftBank, including plans for data centres with combined capacity of 4.5 gigawatts, though financial terms of the datacenter build-outs were not disclosed. Oracle reports that its cloud infrastructure revenue is growing fast (up 77% year-over-year), aided by several large contracts. OpenAI expects about $12.7 billion in revenue this year, and is also said to be working with Broadcom on designing its own AI chip.
  

• Is Senator Ted Cruz another harbinger of the AI apocalypse? Last week, the Texan senator proposed the SANDBOX Act, which would allow AI companies in the US to request exemptions from existing federal regulations for products or services that use AI. These waivers could last two years each, up to a total of ten years. If a regulatory agency doesn’t respond within 90 days, the waiver is automatically granted; if rejected, the company may appeal to the White House’s Office of Science and Technology Policy (OSTP), which can override the agency – a potentially massive dub for Big AI firms.
  

• If you ever wanted greater proof AI is catching up with the times, Google’s own AI search has exhibited its own cognitive dissonance by apparently claiming “DOGE” never existed. Users have found the search tool misinterpreted the term to refer to the Elon Musk-fronted “Department of Governance and Efficiency” – formerly known as the US Digital Services Department – as fictional, nor recognize DOGE as the cryptocurrency Dogecoin. The confusion seems to stem from the AI parsing satirical or political content, which treated “DOGE” as a made-up government department. Needless to say, its own “bizarrely ambiguous” governmental status hasn’t helped either.

  

  Source: Google Gemini via The Verge

• US Senator Elizabeth Warren has written to Defense Secretary Pete Hegseth raising concerns about a $200 million DoD contract awarded to Elon Musk’s AI company xAI for its model Grok. Warren cites issues including Grok’s history of producing offensive content (all heil Mecha Hitler, et al.), misinformation and loose guardrails, with the senator questioning whether Musk and his companies may be improperly benefitting from government access to sensitive data, and a whole raft of anti-competitiveness issues. The lack of irony isn’t wasted on Warren either, who cited criticisms of xAI’s “patchwork” safety fixes and lack of public safety reports for Grok 4.
  

• You would imagine bumholes at Anthropic will be tightly clenched after a federal judge paused Anthropic’s proposed $1.5 billion book piracy settlement last week – another twist in the AI firm’s defense against a class-action lawsuit brought by authors who allege trained its AI models on copyrighted books – some legally purchased, others allegedly pirated. Under the deal, authors and publishers would receive about $3,000 per book, but Judge William Alsup raised concerns that the deal could be “forced down the throats” of authors and their legal representation, who also believe the claims process lacks clarity. Judge Alsup ordered more information and will revisit the settlement in a hearing scheduled for September 25.
  

• In other news likely to infuriate Elon Musk, OpenAI and Microsoft have agreed via a nonbinding memorandum of understanding (MoU) to let OpenAI convert its for-profit arm into a public benefit corporation (PBC), pending regulatory approval. Under the plan, OpenAI’s nonprofit board retains operational control, and the nonprofit will receive a stake in the new PBC worth over US$100 billion. This change could enable OpenAI to raise fresh investment and eventually go public. Microsoft’s blessing is seen as a big win-win for OpenAI, who have been negotiating to loosen the tech firm’s status as a dominant cloud provider while still maintaining an important partnership between the two companies.
  

• A bit of a shitter for Apple fans across Europe: Apple’s new Live Translation feature for AirPods, powered by Apple Intelligence, will not be available in the European Union at launch. The feature, which enables real-time translation during conversations, is supported on AirPods Pro 2, AirPods Pro 3, and AirPods 4. Apple attributes the delay to compliance with the EU’s Digital Markets Act (DMA), specifically its interoperability requirements. The company clarified that data protection regulations are not a factor in this decision.
  

  

  Source: Apple

  🧩 Other Bits
  

• Adoption of AI by large US companies is showing a modest decline, according to recent US Census Bureau data, with many large employers pausing after a rapid initial rollout amid frustration over slow returns on investment. The trend, as discussed by Apollo Academy, suggests that major corporations are now more hesitant, while smaller firms continue adopting AI at a steady pace.
  

• Has Big Government gone too far this time? The Center for the Alignment of AI Alignment Centers (CAAAC) would lead you to believe so, if it wasn’t a very clevel a satirical website poking fun at the often-theoretical world of AI safety and alignment research. CAAAC “does the work literally no one else will”, sending up the proliferation of new “alignment” organizations and hype in the sector. The ‘AGI predictor clock’ is a laugh, too.
  

• ChatGPT VP Nick Turley shared on X that a planned retirement of OpenAI’s Standard Voice Mode will proceed after a 30-day sunset, following community feedback expressing that the feature was meaningful to many users. The company appears to be balancing user sentiment with its product and technical roadmap.

• OpenAI is backing the production of a genAI animated feature film called Critterz, aiming to show that the tech can make movies both faster and cheaper than traditional Hollywood methods. With a budget under $30 million, the film aspires to debut at Cannes and could mark a watershed moment in AI-driven content creation.

• This is one hell of an 180: Google has admitted in a legal filing that the open web is in "rapid decline," despite publicly maintaining that the online ecosystem is thriving and AI isn’t harming web traffic. This acknowledgment contrasts with previous public statements and reflects growing anxieties among digital publishers and regulators about Google’s influence within the world-wide web.
  

• Do you have a shame kink or argument fetish and want to be scolded by your AI? Well for the low, low price of $129 you can now be abused by your own wearable, thanks to a new AI-powered “Friend” necklace that’s designed to constantly listen to user conversations – and then mug you off. That’s a Christmas present for the missus sorted.
  

  

  Source: Friend

📈 Trending tools, models & apps this week

📋 LLM Leaderboard

📲 Trending tools & apps

🫵 Our Picks

• Anthropic just made Claude useful for spreadsheets; you can now edit files, clean data, and handle grunt work inside your docs instead of copying stuff back and forth.

• Genspark AI Browser is the first browser with free, on-device AI. No internet needed, 169 open models to pick from, and it comes with a built-in agent that can hunt down deals or auto-browse the web for you.

• Google’s Veo 3 now pumps out vertical video in 1080p, which means AI-made clips that actually work for TikTok, Reels, or Shorts instead of just looking like research demos.

🚀 Trending Apps & Models

Oboe lets anyone build light, flexible courses on almost any topic just by typing a prompt, think science, pop culture, languages and offers audio, text, visuals, games and interactive tests so learning works the way you want it.

LockedIn lets you drop any job listing URL and the AI will research the role like a human would, then build you custom cover letters, resume tweaks and outreach tips so you stand out without banging your head against the “standard job apps” wall.

OpenAI’s upgraded Codex makes Codex faster, smarter and more capable of running long, complex code tasks on its own, whether you're working in your IDE, terminal, GitHub or even on mobile. It can now review code, refactor at scale and help you catch critical bugs before shipping.

💸 Financials

• Perplexity AI has reportedly raised US$200 million at a $20 billion valuation, bringing its total funding to $1.5 billion and positioning it as a competitor to Google in the AI-powered search market.
  

• Gary Marcus says the math ain’t math-ing with OpenAI, and that the firm has nowhere near the $300 billion sum underpinning the Oracle-OpenAI Stargate deal, claiming we’ve hit “peak bubble” in the sector.
  

• The robotics startup sector is experiencing significant growth, with $6 billion invested in the first seven months of 2025, driven by advancements in hardware, software, and increased investor interest. A new golden age for the industry?
  

• Born, the creator of the virtual pet Pengu, has raised $15 million to launch a new wave of social AI companions, aiming to redefine digital companionship.
  

• Three former Google X engineers have raised $6 million to develop an AI system designed to act as a "second brain," enhancing human memory and decision-making.
  

• Nuclearn has secured $10.5 million to help the nuclear industry embrace AI, aiming to modernize operations and improve safety through advanced technologies.

🕵️‍♂️ FREE ENTRY TO OUR INVITE-ONLY AI CHAT ON TELEGRAM…

If you share this newsletter with a friend and they actively sign up for the Big Machines newsletter, we will send you access to our invite-only Big Machines Telegram group, which is full of builders, investors, founders, and creators.

Access is now only granted to those who refer our newsletter to active subscribers, which means if you sign up on your work email, we will know you sneaky bastards.

This would kill our open rate, so please don't do that, we beg.

👋 Here’s some memes to get you through the week… see you next time

A thought for the week: In our globalised, neo-liberal, technocratic world that seems to be going down the toilet quicker and quicker by the week, please do spare a thought for the billionaires who lose everything.

That’s it til next week, folks. Remember to be kind, love everybody, and do not commit major financial crimes.

Sam, Grant, Mike, Matt and The Big Machines team.

Now, follow us on Twitter.