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: GPT-5 is here and it changes the game, again

The launch of GPT-5 marks a turning point in how we interact with artificial intelligence. OpenAI has replaced multiple earlier models with a unified family – GPT-5, GPT-5 Pro, and GPT-5 Mini – all tied together by a real-time thinking switch that automatically adjusts how much reasoning power to use for any given task. For most people, this means no more guesswork about which model to pick – GPT-5 simply decides and delivers.

But as Ethan Mollick’s early tests revealed, GPT-5 is more than just efficient – it’s proactive. Ask it for a business plan, and it might hand you a landing page, draft copy, financial projections, and a list of next steps without further prompting. Tell it to make it better, and it doesn’t just improve – it adds creative flourishes you never asked for, from subtle visual details in images to complex functionality in code. It’s as if the model has moved from waiting for orders to collaborating with you, suggesting goals as well as executing them.

This automation can be thrilling and a bit unsettling. The same proactivity that makes GPT-5 so powerful also raises questions: whose values and assumptions shape its “next best step”? And in a world where AI can create software, documents, and designs in minutes, the real challenge may shift from building to curating – deciding which outputs are worth keeping. Still, the leap in capability is undeniable. Microsoft has already embedded GPT-5 across its ecosystem, and governments like the US are testing it at scale, signaling that this is not just another model upgrade, but a redefinition of what everyday AI can do.

In short, GPT-5 doesn’t just answer your questions – it runs with them. Whether you find that exhilarating or unnerving might depend on how ready you are for an AI that takes initiative. But one thing is clear: this time, the AI might figure out the next step before you do.

LLMs & AI Models

Cloudflare accuses Perplexity of bypassing site restrictions by masking bots as browsers, affecting tens of thousands of domains.
Microsoft integrated GPT-5 into Copilot, GitHub, and Azure, boosting complex task solving, coding, and agent app creation.
OpenAI offers ChatGPT Enterprise to US federal agencies for $1 in first year, with unlimited use and Deep Research tool.
Google adds Guided Learning mode to Gemini, using interactive lessons to foster critical thinking in students.
OpenAI releases open-source models gpt-oss-120b and gpt-oss-20b, matching or beating prior small models.
Anthropic upgrades Claude Opus to version 4.1, improving code, math, and research accuracy for paid users.
ChatGPT adds mental health breaks, prompting users to pause during long chats and limiting direct advice.
Elon Musk plans to add ads in Grok AI answers to offset revenue decline after leadership change.

New Tools

- Microsoft releases free Copilot 3D, converting 2D images into GLB 3D files for design, prototyping, and learning workflows.

- Google upgrades Google Finance with AI answers, real-time news, advanced charts, and commodity and crypto data.
- xAI launches Grok Imagine, generating 15s videos from text or images, faster than rivals, with free US trial.
- Midjourney adds HD video mode for Pro and Mega users, producing 4x pixel density at 3.2x higher cost.
- Google releases Jules, an async AI code assistant on Gemini 2.5 Pro, with a free tier for 15 daily tasks.
- ElevenLabs debuts Eleven Music, creating multilingual songs with custom lyrics, styles, and structures.
- Google Gemini app gains free Storybook feature to create and narrate personalized children’s books.

Other Quick Picks

- Anthropic identifies “persona vectors” in AI, helping reduce bias, hallucinations, and unexpected behavior shifts.
- OpenAI o3 wins Kaggle AI Chess tournament, beating xAI Grok 4 and exposing Grok’s strategy flaws.
- OpenAI rewards ~1000 staff with multi-million dollar bonuses to retain talent before GPT-5 launch.
- Google DeepMind open-sources Perchmodel for faster, more accurate wildlife sound analysis to save almost extinct birds.

- MIT develops PUPS AI to predict precise protein locations in cells, aiding disease research.
- Trump announces 100% tariff on imported semiconductors to push production to US.
- Universal adds “no AI training” warnings to film credits, aiming to block style replication.
- Nearly 100,000 public ChatGPT chats were indexed by Google, exposing sensitive work and personal information.
- GitHub survey shows developers moving from AI skeptics to strategists, predicting AI will write 90% of code in 2–5 years.

🇪🇪 AI News from EstAIonia

- Delfi test shows popular AI bots can surface Russian propaganda from Pravda in certain queries.
- AI Practitioners Annual Conference (AI praktikute aastakonverents) on Aug 26 in Tallinn focuses on AI adoption in Estonian organizations and strategic implementation. Use code AI2025 to get 50% off, register here.
- DriveX and Swedbank speed up car insurance with AI, guiding clients in photo capture and detecting damages in real time.

🎙️ AIPowerment Podcast is back from summer vacation and dropped a new episode featuring Gerlyn Tiigemäe and Sandra Reivik covering July AI news, new tools, and AI terms.

🎧 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 “Kuidas endale luua personaalne AI assistant?”, 14.08.2025 – information
Keynote Speaker at Äripäev AI Praktikute aastakonverents on the topic of AI for Finance (“Kuidas AI muudab finantsjuhtimist?”) on 26.08.2025