The most important news pieces from the AI world are here for you, whether you’re an AI enthusiast or just wish to keep an eye on what’s going on.
Focus Topic: Microsoft and Google redefine AI interfaces – meet MAI and Nanobanana
The AI world just saw two heavyweight debuts that might reshape how we interact with language and visuals. On one side, Microsoft introduced its first fully homegrown models – MAI-Voice-1 and MAI-1-preview – marking a clear pivot away from full reliance on OpenAI. On the other, Google rolled out its experimental Gemini Flash 2.5 image model, internally known as Nanobanana, which could be a game-changer for creative professionals.
Let’s start with Microsoft. The MAI-Voice-1 model stunned observers by generating a full minute of human-like speech in less than a second – setting a new benchmark for voice AI performance. Meanwhile, MAI-1-preview, their new text model, is designed for efficiency – built to handle typical instruction-following tasks while using far less GPU power than competitors. Reactions from the developer and enterprise communities have been cautiously optimistic. While Microsoft’s move signals growing independence from OpenAI, some experts are waiting to see if these models can scale beyond internal use cases and challenge the likes of GPT-4o or Claude.
On the visual side, Google’s Nanobanana (officially Gemini Flash 2.5 Image) has quietly climbed to the top of the LM Arena image-editing leaderboard. The model enables precise, multi-step image changes, natural-language edits, and visual consistency at a level unmatched by competitors. It supports style transfers, image merging, and even maintains character likeness across edits. Early users are already experimenting with high-end workflows like AI film storyboarding and ad content generation.
Together, these launches signal a shift in how AI tools are built and delivered – from giant all-purpose models to highly specialized, multi-modal assistants built with efficiency and real-world use in mind.
📸 I’ll be sharing a separate LinkedIn post soon with deep-dive visuals and real test results from Nanobanana🍌– stay tuned!
LLMs & AI Models
- OpenAI ended the Realtime API beta and released a new gpt-realtime model with better accuracy, multilingual support, and MCP integrations.
- Cohere introduced Command AI Translate, outperforming GPT-5, DeepSeek-V3, and Google in 23 business languages with secure offline installation.
- xAI unveiled Grok Code Fast 1, a coding model optimized for low-cost agent-based software development tasks.
- Anthropic announced that Claude chats will be used for training unless users opt out before September 28, affecting Free, Pro, and Max plans. Make sure to uncheck that box!
- Anthropic is testing Claude for Chrome, a browser extension enabling autonomous web browsing with security safeguards.
- Perplexity launched Comet Plus, sharing $42.5M in revenue with publishers whose content appears in AI search and Comet browser.
- xAI released the Grok 2.5 open-source model and confirmed Grok 3 will follow within six months.
- OpenAI will add parental controls to ChatGPT after a lawsuit linked the model to a teen’s suicide, exposing long-term safety gaps.
New Tools
- Krea opened a waitlist for Realtime Video, enabling live video creation and editing via drawing, text, or webcam input.
- Microsoft will bring Copilot AI to Samsung’s 2025 TVs and smart monitors as an on-screen assistant for recommendations and Q&A.
- Google launched Gemini Flash 2.5 Image (nanobanana), enabling multi-step, natural language-based edits with strong visual consistency.
- Google upgraded Vids AI with features like image-to-video, AI avatars, and automatic transcript summarization for smoother workflows.
👉 Explore these tools: Krea AI | Google Vids | Gemini Flash 2.5
Other Quick Picks
- Reports show Claude misuse in ransomware, fraud, and extortion schemes, marking a new era in cybercrime.
- TIME100 AI 2025 highlights global leaders shaping the future of artificial intelligence.
- NVIDIA posted $46.7B revenue, up 56% YoY, with $27B from Blackwell chips and record profit of $26.4B.
- Google and rivals narrow ChatGPT’s lead, as Gemini and other apps gain global market share. Here are the 50 most popular AI tools:
- Over 1B people now use AI chatbots like ChatGPT and Gemini regularly, signaling mass-intelligence adoption.
- OpenAI and Anthropic published a joint safety study, testing each other’s models under risky misuse scenarios.
- Melania Trump launched a US AI Challenge for school students to build AI solutions for community problems.
- xAI sued Apple and OpenAI, accusing them of anti-competitive App Store practices favoring ChatGPT.
- Apple is in talks with Google to power Siri with Gemini, as the major Siri update slips to 2026.
- Meta partnered with Midjourney to bring its aesthetics into future visual AI products.
- Claude’s 74,000-chat study showed teachers use AI mostly for admin tasks – 57% for curriculum design, 13% in research, and 7% in grading, which remains most controversial.
🇪🇪 AI News from Estonia
- Äripäev AI practitioners’ annual conference stressed that quality data and clear processes are key to business value creation. I was on stage, talking about AI for finance, and made a summary in my LinkedIn post.
- Telia CEO Andre Visse said the TI-Hüpe program will strengthen Estonia’s competitiveness by bringing long-term benefits to education, economy, and future workforce.
- From 2026, Estonia will implement a major cybersecurity reform, expanding requirements to 6,500 companies and introducing fines up to €10M or 2% of global revenue.
- Estonian software company Trigon adopted a student-built ML model from University of Tartu, achieving over 90% accuracy in invoice accounting suggestions.
🎙️ AIPowerment Podcast dropped a new end-of-month episode featuring Gerlyn and Sandra, covering the latest news, the launch (and slight fail) of GPT-5, the Äripäev AI conference, and some research on AI.
🎧 Listen to AIPowerment Podcast on Spotify, Apple Podcasts, and YouTube.
🙌🏼 If you found this article valuable, subscribe to weekly news updates.
💼 Exploring how to make AI work in finance? Let’s talk practical use cases – connect with Gerlyn Tiigemäe for expert guidance.
📚 If you would like to participate in one of my trainings or listen to speaking engagements, here are the upcoming ones:
- Training “AI võimalused finantsvaldkonnas” @ Äripäeva Akadeemia, 21.05.2025-28.11.2025 – information
- Training “AI võimalused projektijuhtimises” @ Äripäeva Akadeemia, 17.10.2025-28.11.2025 – information
