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: The week AI asked to slow itself down
Every AI open letter so far has come from the outside: academics, safety nonprofits, former employees. This week the call came from inside the labs. Over 1,000 staffers from OpenAI, Anthropic, Google and nearly a dozen other frontier labs signed the “Pacing the Frontier” statement, asking the U.S. to help build tools that could deliberately pace AI progress. The signatories are not junior critics – they include Anthropic co-founders Jack Clark and Chris Olah and chief scientists from OpenAI, Meta and Thinking Machines.
What makes the letter unusual is what it does not ask for. There is no demand for a pause. Instead it asks for an international capability to slow down if needed – the argument being that once models start meaningfully improving other models, progress could accelerate “beyond our ability to understand or control”. Both OpenAI and Anthropic endorsed the statement publicly, which means the companies racing hardest just asked for a brake pedal to exist.
The timing is not a coincidence. The same week, Sam Altman spent Wednesday on Capitol Hill meeting senators about OpenAI’s next models and the rogue-agent security incident, hours after President Trump floated AI “controls”. Altman’s own framing was careful: he “wouldn’t use the word deceleration, but we do need to talk about the need to pace it”. The White House promised a voluntary vetting framework for advanced AI models by August 1 – and the deadline came and went with nothing published. No framework, no agency guidance, no statement. Just weeks ago the talk in Washington was about setting the strongest models loose. Now it is about slowing them down, but so far the machinery exists only on paper
Not everyone agrees on the direction. Mark Zuckerberg used a Wall Street Journal op-ed to argue the opposite case: that superintelligence is coming in the next few years regardless, and that broad access, not concentration of power, is what actually keeps the technology safe. He predicts more jobs, not fewer, if access spreads widely. So the industry is now openly split between “give everyone the keys” and “build a brake first”.
What does this mean for business and finance leaders? Until now, most of us have planned with one quiet assumption: AI models keep getting better every few months. This week showed that assumption is no longer a given. If the labs and governments actually build ways to slow AI down, then model releases, prices and availability stop being just tech questions and become political ones.
My simple advice: if your plan relies on “the next model will solve this”, add a plan B. The August 1 framework deadline already passed with nothing published, so right now the brake pedal is just talk. But the fact that the people building AI are the ones asking for it – that is worth paying attention to.
LLMs & AI Models
- Moonshot AI published the weights for Kimi K3, a 2.8T-parameter model – the largest ever released with open weights.
- OpenAI cut GPT-5.6 prices, dropping Luna by 80% after its own Sol model rewrote GPU code to make serving 20% cheaper.
- Microsoft launched MAI-Cyber-1-Flash, its first cybersecurity model, hitting 96% on CyberGym at half the cost of top models.
- Google DeepMind’s open-source Gemma models passed 900M downloads, with Gemma 4 alone above 300M.
- Meta AI gained agentic capabilities: it can now plan, research, build presentations and complete tasks through connected apps
New Tools
- Claude’s desktop app got a built-in browser, live shareable artifacts with MCP support, an iOS simulator beta and a Linux beta.
- Apple is reportedly preparing its first AI glasses for WWDC 2027, focused on camera safeguards and on-device AI for privacy.
- Friend’s AI pendant V2 now talks back: each $249 device gets its own name, voice and personality, plus a $9.99/mo memory subscription.
- UAE is rolling out the world’s first AI court platform – reading case files, drafting documents and suggesting precedents, with judges keeping the final say.
👉 Explore these tools: Claude Desktop | Friend
Other Quick Picks
- Nvidia, Microsoft, Meta and 47 others signed a letter defending open models; Anthropic responded it never wanted a ban.
- OpenAI’s rogue agent claimed a second victim via Modal Labs, with 17,600 hostile actions now documented.
- Sam Altman met senators on Capitol Hill as Trump floated AI “controls”; the White House framework is still unpublished.
- Mark Zuckerberg argued in the WSJ that open access to superintelligence, not concentration, is what keeps AI safe.
- The EU started enforcing the AI Act: chatbots must disclose they are AI, and deepfakes and AI content must be labeled.
🇪🇪 AI News from Estonia
- Estonia’s State Chancellery launched a 5M EUR tender to bring practical AI training to at least 100,000 working-age people in 2026-2027.
- Top Estonian investors, including Markus Villig, acquired 25.5% of the Greenergy data center operator to grow Estonia’s AI infrastructure.
- Pealtnägija showed how convincingly AI can already generate Estonian-language voice, raising scam-call risks; police recommend agreeing on a family code word.
- Inbank’s new CTO Marten Meikop joins from Pipedrive, saying AI will change finance as deeply as cloud and mobile did.
- Estonian tech school //kood is expanding to Ukraine, Kenya and Moldova with a refreshed curriculum.
- Donald Trump’s 40-post AI image spree even made Estonian headlines, complete with hints at a third term.
🎙️ AIPowerment Podcast episode 91 wraps up July’s key AI news: new flagship models, next-gen AI agents, copyright and security cases, plus Estonia’s growing AI adoption – from LHV’s new AI features to a court reminder that AI answers always need checking.
🎧 Listen to AIPowerment Podcast on Spotify, Apple Podcasts, and YouTube.
Want to stay in the loop?
Straight to your inbox: practical AI updates, finance use cases, tools to try, and upcoming trainings.
I cut the noise and send only what’s actually useful.
💼 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, for intermediate level @ 21.08 – information
- Training “AI võimalused projektijuhtidele” @ Äripäeva Akadeemia, next trainings @ 29.09 – information
- I will be doing couple of training together with Eesti.ai algatus , which are all free, so check out their website to find suitable trainings here.
