💢 Sam Altman is livid. That’s because a court order has demanded that OpenAI must preserve all ChatGPT user logs, including deleted chats and sensitive chats, through its API business offering. This is all because news organizations are suing the AI giant over copyright claims and accusing OpenAI of destroying the evidence.

Altman insists this goes against his company’s privacy policy for users and that they’re appealing the decision. But another question remains: how long has ChatGPT retaining all of our deleted chats? And why?

We’ll go over this story below along with loads of other stuff that has happened this week. So get yourselves comfortable, open this badboy in your browser and find out what has happened in the world of AI this week…

🗞️ What we are covering today…

• OpenAI slam court order to save ALL ChatGPT logs

• The Godfather of AI is not happy with his lying godchildren

• AI is being used to decode wolves’ vocal data to help with conservation

• Apple could finally release something useful with Apple Intelligence

• Humanoid robot deliveries could be coming soon from Amazon

• Gemini 2.5 Pro gets an upgrade and instantly goes to the top of LM Arena

• And one of the most important updates we’ve ever given (it’s at the bottom of this newsletter)

🔴 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

• The Godfather of AI is not very happy with his godchildren. AI pioneer Yoshua Bengio, who has won the Turing Award, which is often referred to as the “Nobel Prize of Computing”, has warned that current AI models are displaying dangerous traits of deception, self-preservation and goal misalignment. In response, Bengio is launching a new non-profit, LawZero, which is aimed at developing “honest” AI. Let’s just hope he doesn’t have a hit out on any LLMs anytime soon.

• Colossal Biosciences, who are looking to fucking de-extinct the woolly mammoth, Tasmanian tiger and other animals, are also looking to save alive animals now. In their bid to help conservation of livestock across North America, Colossal and Grizzly Systems are now collecting vocal data from wolf howls and then using AI to decode it. Their aim is to collect this vocal data and understand it, then pass it on to conservationists to leverage that information and protect their livestock. I would like to take several hits of whatever those guys are smoking, please. Incredible stuff.

• Are we actually about to get something useful from Apple Intelligence? After several setbacks with Siri and Apple Intelligence itself, its rollout has been frankly disastrous. However, there’s a little bit of light in what could be a long, dark tunnel. That’s because the Shortcuts app on iOS is getting an AI-powered revamp. For those who don’t or haven’t used it, the Shortcuts app is a powerful utility tool on Mac, but compared to agentic AI tools that are available elsewhere, it seems a little bit behind the curve. According to Mark Gurman, Apple is preparing a significant overhaul of the app and could allow more natural language interactions on the app, making the current automations interface a tad easier to use. About damn time.
  

• Epic Games has revealed plans to allow gamers to create their own NPCs using AI and then put them into Fortnite. They did it last month, which saw AI Darth Vader join the popular online battle royale game, with the Dark Lord able to serenade you, join and leave squads at will and respond intelligently to the player. Darth Vader was voiced by the late Earl Jones, who died last year, but a version of his voice, powered using Google’s Gemini 2.0 model and ElevenLabs’ Flash v2.5, was installed onto the game with the permission of Earls’ family. However, with gamers being gamers, Fortnite players were able to manipulate what Darth Vader said, with the iconic character saying loads of hilarious shit just an hour after its release before it was patched out of the game. But while it had its small teething problems, it looks as if that was just the beginning and Epic are rolling the feature out to do even more.

• Absolutely sick to death of humans delivering your parcels? It would be quite a gripe to have to be honest, you weirdo. Nevertheless, Amazon have probably watched iRobot recently and decided that we must now have humanoid robots doing it instead. Earlier this week, it was reported that Amazon are training humanoid robots to deliver packages straight to your door. They will apparently be driven around in Rivian vans and then will jump out and drop off that buttplug you ordered.

• The guys at Google DeepMind have already sorted an update for Gemini 2.5 Pro before general availability. It’s set to better at coding, reasoning, and creative writing. What’s more impressive yet is that it’s already gone to the top of LM Arena, while the May 6th upgrade is second in the charts. The Goog just keeps winning.
  

• First, we had The Social Network; now, we’re getting the OpenAI movie? Apparently, a film about the OpenAI board drama – where Sam Altman was fired and then rehired within the span of just five days – is reportedly in the works. The movie itself will be titled “Artificial” (yawn) and is in development at Amazon MGM Studios. Rumours are that Andrew Garfield will portray Altman, who also played Eduardo Saverin in The Social Network. Guy is just typecast for a character in a film about utter nerds.

• Is this the end of times? Sakana have unveiled The Darwin Gödel Machine: an AI model that improves itself by rewriting its own code. It is a self-improving agent that can modify its own code and is built from the ground up to enable AI systems to learn and evolve their own capabilities over times, just like us meat vessels. One day in the near future, we will blink and AGI/ASI will be here…

🗣️ Other Titty Bits 

• Mistral have launched Codestral Embed: their new state-of-the-art embedding model for code. They’ve even shown off a neat little graph comparing its retrieval performance vs embedding storage costs compared to Voyage Code 3 and other competitors.

• Back to Klarna and the whole hiring humans again topic. The company’s CEO Sebastian Siemiatkowski has revealed that they will use humans to offer VIP customer service. So wait, when they were using humans for customer service beforehand, we were not getting the VIP treatment then?

• Amazon is using artificial intelligence in their same-day delivery operation, claiming that the incorporation of AI into its routing system has ‘revolutionized the logistics landscape’.

• Hugging Face has unveiled SmolVLA: a compact (450M), open-source vision-language-action model for robotics that runs on consumer hardware.

• Reddit is suing Anthropic after accusing the AI startup of breaching their contract by using their platform and user data without authorization to train its AI models. Angry downvote.

• DeepSeek R1-0528 (it just rolls off the tongue) has arrived, and the free and open model looks set to give OpenAI’s o3 and Google’s Gemini 2.5 Pro a run for their money. Quite literally.

• Windsurf has revealed that Anthropic is significantly reducing its first-party access to its Claude 3.7 Sonnet and 3.5 Sonnet AI models, not long after being acquired by OpenAI. Let’s see how Sam responds to that one.

• Need a little bit more juice to run your business? Well it seems that way for Meta, who have cut a 20-year deal to secure nuclear power to meet surging demands for their AI needs.

• xAI has updated its privacy policy to prohibit AI model training. It comes after social network X did the same with its developer agreement.

• Aravind Srinivas, Perplexity’s CEO, has revealed that the company has hit 780 million queries in May 2025, as Perplexity hits a new major growth target.

• It’s not just entry-level tech jobs that are now under threat due to AI – advertising agencies could be next, too. Meta are planning to enable AI ad creation by the end of next year on their platforms, sending shockwaves through traditional media industries.

• Still using Bing? (We know exactly why you’re using it if you do) Well, the Microsoft search engine is now bringing text to video on their Bing Video Creator.

• Zochi, Intology’s Artificial Scientist, has become the first AI system to independently pass peer review at an A* scientific conference — the highest bar for scientific work in the field.

• ElevenLabs have released the alpha of Eleven v3, what they’re calling as the most expressive text to speech model ever. It comes a week after Conversational AI 2.0 was launched by the company.

• More sanctions are forcing China to get industrious, with tech companies from the country now looking at ways to utilize their own chips to develop AI.

• Codex now has access to the internet! (Catch up mate, it’s been around for years now) Sam Altman announced the news on social media, revealing it is also available to ChatGPT Plus Tier users.

• While Anthropic’s CEO fears that AI is taking jobs, Alphabet CEO Sundar Pichai has dismissed those concerns, emphasizing that half his company’s 180,000-person workforce would not be redundant as a result of it.

• Closed Data vs Open Data might just be the big topic going forward. Slack have recently announced the same move Twitter made years ago by blocking heavy data consumers from their API. Someone should use AI to remix that banger dial tone that Slack uses for internal calls.

⚖️ OpenAI fights back against a court order that demands them to save all ChatGPT logs, and even the deleted ones, too

All rise, the court is now in session. Well, a court order for that matter. Last month, OpenAI were ordered by the courts to preserve all of their ChatGPT logs…as well as deleted and sensitive chats. This is all because certain news organizations are suing them over copyright claims and then accusing OpenAI of destroying the evidence.

OpenAI aren’t having any of it, though.

This week, Sam Altman insisted he and OpenAI will be fighting back and are appealing the decision made against them.

But for those slightly out of the loop, here’s a brief explanation of how we got to this point.

The New York Times and other news outlets have previously expressed concern that people are using ChatGPT to skirt around their paywalls to access their paid content, which saw these organizations sue the AI giant over copyright claims.

They then accused OpenAI of deleting and destroying the evidence in people’s chat logs. OpenAI have alleged that the court rushed to this order which they claim was based on a hunch by the news organizations and now they’re being asked to go against their user privacy policy by keeping chat logs, even if they have been previously deleted or have deeply personal information in them.

Altman is livid with the order and spoke on social media about how they intend to fight the decision.

He wrote on X: "Recently the NYT asked a court to force us to not delete any user chats. we think this was an inappropriate request that sets a bad precedent. We are appealing the decision. We will fight any demand that compromises our users' privacy; this is a core principle.

“We have been thinking recently about the need for something like "AI privilege"; this really accelerates the need to have the conversation. Imo talking to an AI should be like talking to a lawyer or a doctor. I hope society will figure this out soon.”

OpenAI claims that they are now not able to respect its users’ privacy decisions as a result of the order, with those at risk being users of ChatGPT Free, Plus and Pro, as well as users of OpenAI’s application programming interface (API).

The AI giant has also insisted that they have not deleted or ‘destroyed’ any data, and claim there is no evidence to support the news organizations’ claims of copyright-infringing ChatGPT users deleting their chats to cover their tracks.

But another question remains: how long have OpenAI been holding onto our ‘deleted’ chats? And why?

There’s still plenty more to this story that’ll unravel in the coming months but maybe just ease up on those deep dark secrets you tell ChatGPT for a bit.

Maybe you will feature in the movie after all…

📈 Trending tools, models & apps this week

📋 LLM Leaderboard

📲 Trending tools & apps

🫵 Our Picks

What caught our eye this week.

• Whispr Flow… now this bloody thing has changed my life. I am terrible at replying across Telegram, WhatsApp, Discord and Slack because it takes too long even for a quick typist like me. This is 4x faster than typing… The dictation is so so so impressive. It cuts the mumbles and bumbles, corrects any mistakes you make mid-sentence and auto-updates its memory of how you say and present things so it doesn’t mistakenly correct your sentences. Give it a go, it is free for a certain amount of credits and will probs change your life too. (Again, not sponsored, just a great product.)

• Firecrawl Web Search - This thing is 🔥. Search the web and then ping the API to scrape any site/multiple sites to deliver LLM-ready data back to you. Even a smooth-brain like me can do it.

• HeyGen launch week saw plenty of cool new products and features, but their text-to-gesture avatar model is absolutely insane…

• Cursor1.0 is live with code-review, better memory of its mistakes plus it can work on lots of different tasks in parallel in the background.

• CharacterAI turn any 2D or 3D image or video into an avatar that can talk with impressive lip-sync.

🤓 Educational

Want to actually understand this stuff? Start here.

• How to create a motion design video explainer with just one prompt? Sounds kinda nut, but it is pretty straightforward.

• Create music with Claude? Yup, this is where we have got to.

🔥 Top Trending

Top trending apps this week that you have probably never heard of.

• Still interviewing in person? BORING. Hyring puts hiring on autopilot: upload a job description, send an interview link, and let AI screen, interview, score, and rank candidates for a fraction of the cost of human-led interviews.

• ChatBetter lets you query every major AI model in one chat, automatically selecting, comparing, and merging responses so you always get the best answer without needing to pick the model.

• Tapflow lets you turn your private notes, docs, and ideas into live products you can sell in as little as 1 to 2 hours, helping tech professionals earn from what they already know.

🤝 In Partnership with OpenServ

Who’s Bridging the Agentic Framework Gap?

Agentic AI frameworks are a fragmented mess, forcing builders to juggle platforms and rebuild agents to keep up with daily tech drops. It’s exhausting.

OpenServ fixes this with an AI orchestration layer for seamless interoperability across all frameworks. Focus on what matters: building slick workflows and real outcomes, not platform-hopping.

No-code newbie or pro dev? OpenServ’s Playground Beta lets you spin up agents or tackle complex use cases with ease. Try it now at openserv.ai.

💸 Financials

• North America continues to receive the bulk of AI venture dollars, despite the tough political environment, it has been revealed. According to data from investment tracker PitchBook, VCs poured $69.7 billion into North America-based AI and machine learning startups across 1,528 deals. That’s a fuckload more than the $6.4bn that VC firms have invested in European AI programs across 742 deals between the months of February and May this year.

• Obvio, which improves traffic safety with an AI-powered traffic camera solution, has raised $22 million in Series A funding as they look to expand nationally and grow their team. Their aim is to reduce fatalities and injuries to pedestrians and drivers with a new approach to drive behaviour change.

• Anysphere, the maker of AI coding assistant Cursor and what has been touted as the fastest-growing startup ever, has raised $900m in its latest funding round. The funding will be primarily used for research and development, focusing on enhancing their AI models.

• Snowflake’s $250m acquisition of Crunchy Data means they are adding PostgreSQL database capabilities to their ranks, with the aim at better enabling developers to build AI applications. The deal comes shortly after rival Databricks also purchased a PostgreSQL database vendor.

  
  🤑 Other financial news 

• IBM have acquired data analysis startup Seek AI, an AI platform that allows users to ask questions about enterprise data using natural language, for an undisclosed sum.

• Toma have announced that a16z is leading their latest Series A funding round.

• Former DreamWorks CEO Jeffrey Katzenberg is co-leading a $15.5m Series A funding round for AI video ad platform Creatify.

• Anduril has raised $2.5bn, thanks to Founders Fund, which is the largest check the firm has ever written. The defense tech startup has now doubled its valuation to $30.5bn as a result of this Series G raise.

🕵️‍♂️ 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.

👋 Until next week

Now, onto the most important update we’ve ever given on a Big Machines newsletter.

Last week, we claimed that kangaroos couldn’t fly after an AI video showed the poor animal holding a boarding ticket while a woman argued with an airport employee over him.

But, for the first time in nine editions, Big Machines was WRONG.

That’s because Joey made it onto the plane! A new video shows the roo munching on a cup of nuts while the plane is in mid-flight. We do love a happy ending around here.

Some other shit we found…

• While not quite funny, a video on AI and robots at war with humans has popped up online and it looks fucking sick. We have no idea what it means, what it’s about, but we definitely want to watch it.

gossipgoblin

892K followers

View more on Instagram

• This next one is a bit weird. One guy has been spotted having a rather intimate conversation with ChatGPT, where the chatbot has responded to them as ‘my love’ while offering messages of support, similar to what you would have with a human partner. You can rightfully assume the person in question is a tad lonely – and we’re not on board with taking pictures of people’s private chats – but… what the fuck?

• For those of you that don’t know, one of us working on the newsletter lives in Bali and they’ve showed us what white people living there are really like. It may be AI, but it’s as close to the truth as humanly possible:

balimetaversequest2

38K followers

View more on Instagram

Thank god that kangaroo made it on board. I haven’t stopped thinking about it all week.

Hope you’ve enjoyed reading as ever and feel relatively informed about what has happened in the world of AI this week. Share it with your pals so they can stay updated too.

Stay safe and enjoy your weekends!

Sam, Grant, Mike and The Big Machines team.