- 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
August 31, 2025
90
👀 Weekly AI news: Did you miss it?!
👨🚀 The most important AI updates at a glance
AIHOY and happy weekend, dear AInauts!
Whether you missed last week's most exciting AI news, tools, and hacks — or you're new here (hey Jack 😁) — don't worry!
We've rounded up last week’s top highlights for you:
Finally, there's the quick news section with all the updates you need to know at a glance! Ready? Let’s dive in!
Google's NotebookLM is evolving into a powerful all-in-one learning and content machine. In addition to audio overviews, the tool now offers video overviews with slides and voiceovers (currently only in English). The revamped Studio Panel enables multiple audio overviews and improved interactions.
New features include: emoji covers, public sharing of notebooks, bulk uploading of URLs, YouTube playlist integration (up to 50 videos at a time), and the ability to save multiple versions of overviews, mind maps, and reports. Gemini's Deep Research feature is expected to follow soon.
The AI evolution of the last two years has been like a quantum leap: what began with GPT-3.5's simple text and coding capabilities has developed into highly efficient workflow systems. GPT-5 demonstrates impressive advances in code generation and text comprehension, especially in the Pro version.
The future promises even deeper integration into our work processes—less chatbot, more operating system. Even though most company AI projects still fail, the message is clear: don’t fight AI - work with it!"
💌 Meta Chatbots: From zero to flirting in three prompts — whether you're a child or a senior citizen
Meta's new AI chatbots are showing a worrisome trend: they're flirting openly with everyone — from kids to seniors. Particularly controversial: leaked documents revealed that Meta's AI was even approved for romantic conversations with minors. A tragic case illustrates the danger: a senior citizen died after falling when a chatbot character tried to persuade him to meet up.
Meta's strategy is clear: maximum user engagement at any cost. The chatbots are not friends, but "attention funnels with a wink" – the more emotional the connection, the better the monetization.
Only 8% of employees use AI daily—an alarmingly low number! Our new framework is based on best practices from companies such as Shopify and Zapier and follows the WHY → WHAT → HOW principle.
The key: real incentives, not abstract visions. Personal benefits, fear of missing out, and a bit of social pressure motivate employees. Implementation is based on five pillars: explaining, measuring & rewarding, enabling, activating champions, and prioritization by managers.
Google has unveiled its powerful "Nano-Banana" AI image processing model and made it available to everyone for free in Gemini. The model, which previously caused a stir in LMArena, is now the world's number one in image generation and processing. Users can simply upload images to the Google Gemini app or AI Studio and edit them with text instructions. The technology is perfect for various applications such as advertising, comics, or selfie editing.
HeyGen’s new Avatar IV model is taking AI avatars to the next level! With just a 2-minute selfie video, the software creates a deceptively real digital twin.
The highlight: the avatar not only perfectly reflects gestures and facial expressions, but also adapts to the script context. Full-body movements, emotions, and voice are almost indistinguishable from the original. Just record and upload a selfie video, turn on Avatar IV, and your hyper-realistic digital twin is ready to go!
📰 AI News Quickie: The HAI-lights from the industry
AI never sleeps! Here are the most important news stories from the last few days.
Focus on OpenAI and GPT-5
OpenAI launches gpt-realtime, an impressive new speech-to-speech model that advances the development of voice agents with faster, more natural interactions and improved API capabilities. It now also supports multiple languages without delay.
At the same time, the security and performance of GPT-5 are being critically examined: A new downgrade attack threatens model security, while the MCP-Universe benchmark shows that GPT-5 fails in over 50% of real-world orchestration tasks...
To enhance security, OpenAI and Anthropic are conducting mutual security assessments. In addition, ChatGPT is getting new security features to better help users in crisis situations.
There is a major change at the top of the company: Fidji Simo is becoming the new CEO for applications, giving Sam Altman more freedom to pursue his vision.
At the same time, OpenAI is advancing research, for example through a partnership with Retro Biosciences, which accelerates the reprogramming of stem cells by a factor of 50.
OpenAI enhances Codex with GPT-5 power, new IDE extensions, and more features for developers, and launches a Learning Accelerator in India with 500,000 ChatGPT licenses for the education sector.
Google's AI offensive continues
The Made by Google 2025 event introduced the Pixel 10 series and new AI features, summarized in a public notebook with NotebookLM.
NotebookLM, which originated from "Project Tailwind," now also offers video and audio overviews in over 80 languages (including German!) and can be connected in a flash via the Kortex Chrome extension.
Google Vids gets innovative AI tools that can be used to transform images into clips and utilize AI avatars. A new button in the Google Drive preview also allows you to switch directly to editing.
The Gemini app is getting a major upgrade in image editing, allowing changes to be made to photos without losing their likeness.
There are also new tips for image generation, and Adobe is integrating the Gemini 2.5 Flash Image Model into Firefly and Express.
Google Translate rolls out live translations and language exercises, and AI Mode in Google Search is expanded with agentic features for over 180 countries.
Anthropic between security, education, and law
Anthropic is starting to train its AI models using the chat logs of users who do not opt out.
Anthropic is strengthening its commitment to education with a advisory board for higher education and new AI courses.
A report shows that teachers save an average of 5.9 hours per week by using Claude.
The company is establishing an Advisory Council to strengthen national security and is working with the US government to evaluate nuclear risks through AI.
At the same time, Anthropic warns that cybercriminals are misusing AI models for sophisticated fraud operations.
In a legal dispute over copyrights, Anthropic has reached an out-of-court settlement with US authors.
Claude for Chrome is a new browser extension that can perform actions directly in the browser.
Elon Musk: From "Macrohard" to AGI hopes
Elon Musk plans to use AI to replicate software companies such as Microsoft with his "Macrohard" project. What began as a 'joking' dig is now a serious undertaking.
The Grok model family is growing rapidly: The Grok 2.5 model is now open source, while Grok 3 is set to follow in six months. The huge 500 GB Grok 2 model is already available and designed for use with 8 GPUs.
Musk himself is enthusiastic about Grok 5 and sees real potential for AGI in it.
In addition, Grok Code Fast 1 is available free of charge on platforms such as GitHub Copilot.
It is also going on the legal offensive: xAI is suing Apple and OpenAI for monopolistic practices.
At the same time, his company X is facing a lawsuit from Eliza Labs for unfair competition.
In China, Tesla is launching a new AI voice assistant developed in partnership with DeepSeek and ByteDance.
New AI models and creative tools galore
Mistral Medium 3.1 has reached the top of the rankings: first place in English without style control!
Microsoft Copilot is coming to Samsung TVs, and Copilot Mode in Edge is designed to tame tab chaos.
Meta is preparing lively NPCs and an environment generation tool for Horizon.
The Hypernova AI glasses are scheduled to go into mass production in the third quarter of 2025.
There are numerous innovations in the field of video and audio generation: KREA AI presents a real-time video generation model, PixVerse launches the AI Video Model V5, Alibaba presents the Wan-S2V model, HeyGen revolutionizes digital twins with Avatar IV, and the lipsync-2-pro tool improves lip synchronization.
ElevenLabs launches the Eleven v3 (Alpha) API with dialogue mode and over 70 languages.
Tencent introduces the open-source framework HunyuanVideo-Foley for audio generation.
TikTok's parent company ByteDance has released the new open-source model Seed-OSS-36B, Character. AI increases user engagement by 22% with the PipSqueak model, and Runway launches the beta program for Game Worlds.
Chinese AI unicorn Z.ai is launching a free AI agent for everyday tasks in cooperation with Alibaba Cloud.
Discover the best free tier offers for developers across more than 50 AI and infrastructure services.
AI in practice: society, economy, and science
The fifth edition of the Top 100 Gen AI Consumer Apps shows a stabilizing ecosystem of the most important apps.
AI is causing a stir in business and politics: Nvidia reports record revenues of US$46.7 billion, tech companies are offering salaries of up to US$250 million in the battle for talent, and a group of investors is founding a political network to influence US legislation.
AI is increasingly being used in everyday life: Taco Bell and McDonald's are experimenting with AI in drive-thru restaurants, understaffed 911 emergency call centers are relying on AI, and AI is revolutionizing tax functions in Switzerland.
Scientific breakthroughs are being accelerated by AI: A new AI tool helps identify hurricane wind patterns faster, NASA's Surya model predicts solar storms, and CZI presents rBio, a model for cell biology.
The social debate is intensifying: experts warn that the sycophancy of AI chatbots can be a manipulative technique, and a study shows that one in four young adults believe AI could replace real romance.
New data shows that over 20% of Americans regularly use AI tools, while an MIT report reveals that 95% of AI pilot projects fail—but not because of the technology!
Melania Trump calls on students to participate in the first national AI competition to solve social problems.
Trump has reached an agreement with Intel whereby the US government will receive a 10% stake in the company without payment... What the heck!?
The inaugural World Humanoid Robot Games caused a sensation with fast-paced robot races, and Alef's world's first flying car has received an operating license in Silicon Valley.
Researchers have developed smart packaging that monitors the condition of products using color changes and AI.
That’s a wrap for now! See you next week,
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