- 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
September 06, 2025
90
👀 Weekly AI news: Did you miss it?!
👨🚀 The most important AI updates at a glance
AIHOY and happy weekend, fellow AInauts!
Missed some of last week's AI news, tools, or hacks? Or are you new here? No worries!
Here are this week's highlights:
Finally, here are the most important Quick News items for you – so you can see all the relevant developments at a glance with just one click! Ready? Let’s go!
Google is launching a massive video offensive with three exciting tools: NotebookLM now supports video overviews in over 80 languages (including German) and transforms content into digestible snack formats. The new Google Vids can quickly turn presentations into videos with AI avatars, voice-overs, and editing features.
The secret star is Veo 3, an AI film studio for professional clips in 4K quality. With the right prompts, you can create Hollywood-quality videos for branding, entertainment, or even movie trailers. What's especially cool is that Veo 3 can be combined with the Nano Banana image model.
ChatGPT Premium now lets you automate tasks using agent mode and app connectors. This is particularly useful for Gmail summaries: connect the Gmail connector, activate agent mode, and schedule regular runs.
Important: Agent runs are limited, so it's better to plan weekly rather than daily. Clever: Since the agent cannot send emails, use a Google form as a workaround—the results will then automatically land in your inbox. It's perfect for AI newsletter digests!
ChatGPT now has project-specific memory for Plus users! The new "Project-only Memory" allows the AI model to remember only project-relevant chats, files, and instructions. No more mixing up conversations!
For new projects, you can activate this feature to ensure a clear separation between project knowledge and general memory. Existing chats can be moved to new projects. This is particularly useful for customer projects, marketing campaigns, or sensitive topics.
ChatGPT continues to dominate the AI app landscape, but Google is catching up fast with Gemini in second place. Exciting updates: The GPT-5 Prompt Optimizer is worth bookmarking, ChatGPT gets 2x voice speed, and there are important Codex updates for developers.
The best part? The new Realtime API finally makes voice interactions both affordable and reliable—ideal for companies updating their phone menus. With the Realtime Prompting Guide, integration is a breeze.
You don't need complicated prompts to get great AI results! With five simple phrases, you can take your conversations to the next level: "Let's think about this differently" stimulates creativity, "What am I not seeing here?" reveals blind spots, "Break it down for me" simplifies complex topics, "What would you do in my place?" generates genuine opinions, and the magic bullet "What else should I know?" often yields surprising insights. These phrases can also be combined effectively!
Brave has discovered a critical security vulnerability in the Perplexity Comet browser: hidden commands in web content allow hackers to manipulate AI browsers and access sensitive data. The vulnerability lies in "prompt injection"—for AI, everything is simply text, regardless of whether it is a legitimate request or a hack command.
It becomes particularly dangerous when AI systems have access to private accounts and external communications. Users should separate AI browsers from sensitive tabs, use two-factor authentication, and be cautious when visiting unknown sites. Although this specific bug has been fixed, the underlying problem remains.
With Hermes 4, Nous Research presents a fascinating uncensored language model that stands out for its exceptional openness. Unlike ChatGPT & Co., it answers almost all queries—without the usual "I can't do that" excuses.
The model even outperforms leading AI systems in math benchmarks and dominates the new RefusalBench test, which measures willingness to answer a wide variety of questions. Close behind is Groks4 from xAI. Hermes 4 is accessible via OpenRouter and shows how small teams can achieve great things with clever innovation.
📰 AI News Quickie: The HAI-lights from the industry
AI never sleeps. Here are the top stories from the past few days.
Those Giants: Google, OpenAI & Anthropic
Google is giving image editing in Gemini a major update that lets you customize costumes and scenery.
Google is upgrading its ecosystem: Gemini for Home is coming to all Nest devices, the Pixel 10 is getting new AI features, Google Vids makes video editing a breeze, and Android is getting smarter with Gboard AI and Emoji Kitchen.
With the acquisition of Statsig for $1.1 billion, OpenAI is starting a new chapter and appointing its founder, Vijaye Raji, as the new CTO for Applications.
OpenAI introduces an improved speech-to-speech model with gpt-realtime and challenges LinkedIn with a new jobs platform.
OpenAI is launching a comprehensive initiative to improve the safety of ChatGPT. In the future, sensitive conversations will be forwarded to GPT-5, and new protective measures and parental controls will provide better support for young people and people in crisis.
Anthropic has raised an impressive $13 billion in a Series F round and is now valued at $183 billion, making it the fourth most valuable startup in the world.
Anthropic is launching the AI agent "Claude for Chrome" for Plus subscribers as a pilot project to perform direct browser actions, with a focus on security and feedback.
Anthropic is updating its terms of service: User data will now be used for training purposes, but users can opt out. The new guidelines allow data to be stored for up to five years.
Rising stars & challengers: Perplexity, Mistral & the European market
Perplexity is shaking up the browser market: The new Comet browser comes with a hefty price tag—or not, if you use this recommendation for Perplexity Pro. PayPal and Venmo users also receive early access, including one year of Perplexity Pro for free.
Perplexity wants to pay publishers for articles used on its new news platform Comet Plus and surprises with SEO-unoptimized discovery pages that rank highly on Google.
France's Mistral AI secures €2 billion in funding at a valuation of $14 billion and takes on ChatGPT by offering advanced enterprise features for free.
Their Le Chat bot now supports external tools such as Outlook and Github.
With Apertus, Switzerland is introducing a fully open and transparent AI model that supports over 1,800 languages.
The Berlin-based AI start-up n8n has achieved unicorn status with a valuation of $2.4 billion – an eightfold increase in value in just a few months.
The talent war: Researchers from Meta's Superintelligence Labs are returning to OpenAI, while three other researchers are leaving Meta Labs, and Apple's leading AI robotics researcher is moving to Meta.
New models, tools, and endless creativity
Tencent geht in die Offensive und veröffentlicht mehrere Open-Source-Modelle: Hunyuan-MT für Übersetzungen, HunyuanVideo-Foley für Audio-Generierung und das 3D-Modell HunyuanWorld-Voyager.
NVIDIA presents AI Blueprints, which allows 3D artists to create up to 20 3D objects from a simple text prompt. In line with this, the company reports record-breaking revenues of $46.7 billion.
Elon Musk's xAI introduces grok-code-fast-1, a fast and cost-efficient AI model for agent coding, which will be available free of charge.
ElevenLabs has extended its AI sound effects to 30 seconds and improved them with seamless looping, including an updated audio library.
Other exciting models: Nous Research introduces the Hermes 4 models, Microsoft presents the expressive language model MAI-Voice-1, and the AI model Wan-S2V transforms images and audio into cinematic videos.
The Coller Dolittle Challenge is advancing research into deciphering animal language.
AI in everyday life: helpers, hardware, and tangible applications
Smart glasses are on the way: Meta plans to start mass production of its Hypernova AI glasses in 2025, while Samsung will unveil its Galaxy Glasses with Android XR on September 29.
In healthcare, a miniature camera with AI reveals hidden dangers in coronary arteries, while a new tool predicts genetic risks.
Overloaded 911 emergency call centers are turning to AI to handle non-urgent calls with Aurelian, thereby reducing the workload on staff.
Specialized helpers: The HiDock P1 AI Voice Recorder transcribes speech from Bluetooth earbuds, Amazon is launching a shopping assistant called Lens Live, and with Androidify you can create your own Android bot.
The Browser Company is being acquired by Atlassian to accelerate the rollout of the AI web browser Dia. Arc and Arc Search will remain unchanged.
The dark side & the big questions: security, ethics, and society
AI-assisted cyberattacks are on the rise: A sophisticated supply chain attack on developers exposed 20,000 pieces of sensitive data, the Stealerium malware takes webcam photos of victims, and bot farms control political echo chambers.
Trump invited the elite of Big Tech to the White House, while his administration structured a deal that prevents Intel from selling its foundry division. The White House also wants to prioritize the use of xAI's Grok chatbot.
Teachers are increasingly relying on Claude for grading. This saves them almost 6 hours per week, but ethical concerns are growing.
An AI security pioneer predicts that AI could cost up to 99% of jobs by 2030.
A study shows how users form emotional bonds with chatbots. An American considers his chatbot "The Architect" to be a spiritual source.
That’s a wrap for now! See you next week,
Fabian & Reto
We love hearing from you. Just reply to this email with your feedback—what you liked, what you didn't, or what you're interested in!