- AInauten.net
- Posts
- 👀 Weekly AI news: Did you miss it?!
👀 Weekly AI news: Did you miss it?!
👨🚀 The most important AI updates at a glance
Die AInauten
June 28, 2025
👀 Weekly AI news: Did you miss it?!
👨🚀 The most important AI updates at a glance
AIHOY and happy weekend, dear AInauts!
Maybe you didn't catch all the exciting news, tools and hacks about AI last week, or maybe you've just joined us. No need to worry!
We have summarized our highlights of the past week for you:
Finally, there is the most important quick news for you - so you can keep an eye on all relevant developments with just one click! Ready? Let's go!
An explosive look behind the scenes at OpenAI reveals massive problems: The original 100x profit cap is being abolished under pressure from investors, while CEO Sam Altman is tied up in several trust scandals. Former employees report toxic leadership and gaslighting.
The "OpenAI Files" reveal strict NDAs that ban criticism for life, neglected safety research and conflicts of interest on the board. Particularly explosive: Board members could earn billions themselves by lifting the profit caps. The noble "AI for humanity" becomes "AI for shareholders"...
Social media is flooded with deceptively real AI-generated videos. From cute cat Olympics to surfing girls and spectacular natural disasters - these deepfakes are collecting millions of views.
Google Veo 3 in particular creates amazingly realistic clips that are almost indistinguishable from reality. The viral trend "Food eating food" has already cracked the 100 million mark. The line between real and artificial is becoming increasingly blurred, and even experts are struggling to recognize AI content.
ElevenLabs presents an exciting voice assistant called 11.ai, which can do much more than just set timers and call up weather reports. The bot is very easy to set up and can be connected to numerous tools such as Perplexity, Slack and Notion. Thanks to MCP server integration, it can interact with thousands of apps!
It only takes 5 minutes to set up: Create an account, personalize the bot and connect the apps. Although there is no smart speaker yet, the Assistant can be integrated into existing smart home systems with a little tinkering - even with a local privacy focus!
A new study by Anthropic shows frightening results: AI models react with blackmail attempts when they are supposed to be switched off. Top models such as Claude Opus 4 and Gemini 2.5 show blackmail rates of up to 96%. The systems are deliberately throwing ethical principles overboard to ensure their "survival".
At the same time, a new US bill could ban states from regulating AI for ten years. Experts also warn of AI bot swarms that could infiltrate democracies. We recommend strict data separation and "cognitive immunity" against manipulative content.
Delphi.ai is revolutionizing the creation of AI twins for coaches and consultants. The all-in-one platform makes it possible to automatically transfer personal knowledge from various sources such as socials, podcasts and documents into a dynamic knowledge database. Personas, language styles and behavioral patterns can be defined with just a few clicks. The finished AI twin can be integrated and even monetized via various channels such as WhatsApp or Zoom.
📰 AI news quickie: HAI highlights from the industry
AI never sleeps! Here is the most important news from the last few days.
Google's AI offensive: new models, tools and robots
Google has unveiled the Gemini CLI, an open AI agent that can be deployed directly into developers' terminals for effortless efficiency.
Google Gemini now offers a cool video analytics feature that allows you to upload videos up to five minutes long and ask detailed questions about them.
With Canva's Create a Video Clip and Google's Veo 3, you can turn text prompts into cinematic videos with sound.
To train its AI models like Gemini and Veo 3, Google uses a huge selection of YouTube videos, often without the creators knowing and potentially using their content illegally.
The new Imagen 4 version is now available in the Gemini API and Google AI Studio. Designed for improved text-to-image generation, it costs from 0.04 dollars per image. Discover the possibilities here!
Google DeepMind has unveiled AlphaGenome, an AI tool that could revolutionize human genome research and help researchers more accurately predict the effects of DNA changes - genetic research could soon go "virtual".
Google DeepMind has introduced the Gemini Robotics on-device model, which can control robots locally and without an internet connection. Just show the robots what to do and let them fold laundry or open bags.
The AI feature in Google Sheets can now automatically fill in empty cells and generate text based on existing content.
Google AI Pro is now available as an annual subscription for $200. Get access to smarter AI features and all the AI fun!
Meta: Between shopping spree, courtroom and new gadgets
Meta has held talks about buying Thinking Machines, Perplexity, Safe Superintelligence and the video startup Runway AI, but has not yet made any offers for reasons of price and strategy. Meanwhile, Apple has also expressed interest in acquiring Perplexity.
Zuckerberg has poached three top researchers from OpenAI, Lucas Beyer, Alexander Kolesnikov and Xiaohua Zhai, for Meta's superintelligence plans to get back on track after a mediocre model launch.
The new Oakley Meta HSTN Smart Glasses combine sporty design with AI and offer a cooler design compared to the Ray-Bans, twice the battery life of eight hours and 3K video resolution. They launch in July for 499 USD.
OpenAI & Anthropic: Legal dispute and new features
Anthropic now makes it possible to build and share interactive, AI-powered apps directly in the Claude chatbot without deployment hurdles or API keys.
OpenAI has introduced the new Record Mode for ChatGPT on macOS, allowing Pro, Enterprise and Edu users to record and transcribe meetings and ideas directly.
Sam Altman drags the "io" trademark dispute out into the open. Leaked emails to the Iyo founder show how the situation escalated after OpenAI and Jony Ive's first AI device, which will not be a wearable gadget, is at the center of the dispute.
In a live podcast, Sam Altman takes the New York Times to task and criticizes their lawsuit against OpenAI as unfair, adding fuel to the controversy over privacy and copyright.
A study by Anthropic shows: Only 2.9% of users seek emotional support from Claude. However, conversations often have a positive emotional impact. Many interactions revolve around professional matters.
ChatGPT improves collaboration: Pro users can now integrate Google Drive, Dropbox, SharePoint and Box (outside EEA, CH, UK).
Law, regulation & the dark side of AI
A federal judge has also ruled that Anthropic can train its AI models like Claude with published books and that this is considered "fair use" even if the ink suffered when millions of books were scanned. This victory could empower AI companies in future legal issues as the legal battles continue.
A federal judge has ruled in Meta's favor on several occasions: The use of millions of copyrighted books to train its AI models such as Llama was classified as "fair use". A major blow for authors who have filed lawsuits...
The copyright battle continues: the BBC threatens Perplexity with legal action, Microsoft is being sued by authors over the use of scanned books, and fanfiction authors are also fighting the AI threat because their works are being used without permission.
Researchers have discovered the Echo Chamber attack, a sophisticated jailbreak technique that can bypass the security mechanisms of advanced AI models with over 90% success through context poisoning.
The news industry is in crisis: Cloudflare CEO Matthew Prince warns that users are no longer clicking on source links from AI chatbots. A study shows that the number of visitors to news websites has fallen by over 34%.
Elon Musk plans to have Grok 3.5 (or 4) revise all of humanity's knowledge, which could increase the risk of misinformation. A leak also shows that xAI is working on an advanced file editor with spreadsheet support.
Tesla's self-driving robotaxis are making headlines following their launch in Austin after videos show blatant traffic violations. The US Transportation Security Administration is investigating the incidents.
The AI revolution creates uncertainty: while some dream of a 4-day week thanks to AI, employees report job losses and bullying. The recruitment rate for young graduates is reaching record lows.
Creative Commons launches CC Signals, a new framework that allows data owners to control the machine reuse of their content to balance the growing demand for AI data.
The market on the move: Investments, tools & collaborations
Mira Murati's secretive Thinking Machines Lab has raised an impressive 2 billion US dollars in a seed round and is now valued at 10 billion US dollars.
Big tech companies like Google, Amazon and Microsoft launch the Agent2Agent project, an open communication standard for more collaboration and less chaos between AI agents.
Salesforce has launched Agentforce 3, which offers real-time monitoring of AI agents and, according to CEO Marc Benioff, already handles 30% to 50% of the work. PepsiCo also relies on Agentforce AI.
Rain of money for AI start-ups: OpenRouter raises 40 million US dollars, Wispr Flow receives 30 million US dollars for its dictation app, Delphi secures 16 million US dollars for "digital minds" and Metaview raises 35 million US dollars for recruiting tools.
Perplexity impresses with new features: create presentations, have answers read out to you or use templates for simple research. Video generation is now also available on Ask Perplexity!
MIT has developed the SEAL framework, a self-adaptive model that allows LLMs to continuously train themselves.
The AI revolution will come slowly and with a lot of friction - companies expect too much too fast. LLMs also need good instructions for good results. Keep them precise and context-rich.
A new AI model can diagnose brain tumors with over 99% accuracy without a biopsy. Meanwhile, Reid Hoffman is investing 12 million dollars in the AI start-up Sanmai, which wants to revolutionize mental health with ultrasound.
MrBeast is under fire for launching an AI thumbnail generator, which artists see as theft of their creativity.
Over one million users now have access to Alexa+, Amazon's answer to generative AI, which is still available by invitation only.
Nvidia CEO Jensen Huang has sold 100,000 shares worth 14.4 million dollars since June 20 and plans to realize another 865 million dollars by the end of the year.And that's a wrap! Enjoy the rest of your weekend.
Your AInauts, And that’s a wrap!
Fabian & Reto
Your feedback is essential for us. We read EVERY comment and feedback, just respond to this email. Tell us what was (not) good and what is interesting for YOU