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 real AI shift - from intelligence to execution risk
For the past years, the conversation around AI has been dominated by one question – how smart can these systems get? But this week highlights a more important shift: AI is no longer just producing outputs, it is starting to act.
We are seeing models and systems that don’t just answer questions, but execute workflows, interact with environments, and even identify vulnerabilities. The discussion around Claude Mythos and its potential ability to discover zero-day exploits shows how far this has gone – AI is entering domains where actions have real-world consequences.
At the same time, companies are beginning to train AI on real human behavior at scale. Meta’s initiative to collect employee activity data to improve AI agents signals a move toward systems that learn not just from text, but from actual work patterns.
This creates a new layer of risk. The question is no longer just whether AI is correct, but whether it should act at all – and under what constraints. Because once execution is automated, mistakes are no longer isolated outputs, but scalable actions.
The next phase of AI will not be defined by intelligence alone, but by how well we manage control, responsibility, and the boundaries of autonomous execution.
LLMs & AI Models
- OpenAI launched GPT-5.5, matching top reasoning models while improving efficiency and pricing for large-scale use.
- ChatGPT introduced Workspace Agents, enabling teams to automate multi-step workflows with memory, tools, and permissions across apps.
- Anthropic expanded its Amazon partnership, securing massive compute power and billions in funding to scale Claude globally.
- Anthropic fixed Claude Code issues, resolving context loss and response limits that impacted output quality.
- Claude Mythos reportedly leaked, exposing risks of controlling access to highly capable restricted AI models.
- Google launched Deep Research agents, generating structured reports from web, files, and private data sources.
- Moonshot AI released Kimi K2.6, outperforming top models in reasoning and coding at significantly lower cost.
- OpenAI introduced Chronicle, using background agents to record workflows and build persistent memory.
- Microsoft made Copilot agentic, enabling multi-step actions directly inside Word, Excel, and PowerPoint.
New Tools
- Anthropic launched Claude Design, turning prompts, screenshots, and code into interactive prototypes and assets.
- Ideogram launched Custom Models, letting users train image generation on 15–100 visuals for consistent brand style.
- Google open-sourced DESIGN.md, enabling AI agents to understand brand rules like colors and accessibility.
- Genspark introduced Build, an agent that creates apps and websites directly from text descriptions.
- Adobe released CX Enterprise, coordinating marketing, content, and customer workflows with AI agents.
- Claude added Live Artifacts, enabling real-time dashboards and auto-updating outputs with version history.
- OpenAI upgraded ChatGPT Images 2.0, adding planning, validation, 2K resolution, and multi-image generation.
👉 Explore these tools: Ideogram | Genspark | Claude Design
Other Quick Picks
- The White House accused Chinese labs of large-scale AI model distillation, raising concerns over security and innovation origins.
- An Anthropic study found heavy AI users fear job loss 3x more, especially among engineers and early-career workers.
- Meta plans to cut 10% of staff, citing AI-driven efficiency and shifting resource priorities.
- SpaceX partnered with Cursor, investing billions to enter the AI coding space and compete with major players.
- Meta is collecting employee activity data to train AI agents, raising concerns around privacy and consent.
- Deezer reports 44% of new music is AI-generated, with most flagged as fraudulent and removed from recommendations.
🇪🇪 AI News from Estonia
- Äripäev is organizing their annual AI conference, which seems to be really interesting because different AI experts (including me!) will be covering different parts of the AI implementations process. If you wish to go, use “AIFinance” for proper discount 🙂
- AI tools can multiply productivity, with experts noting small monthly costs can generate thousands in value for businesses.
- Claude Mythos raises cybersecurity concerns in Estonia, with potential to discover vulnerabilities and push need for stronger defenses.
- The government is investing €6M+ into deep tech R&D, supporting AI, data science, and innovation-driven companies.
- An EY study shows 84% have used AI, while 16% trust it with decisions, but concerns over safety and fake content remain high.
- NATO’s Locked Shields cyber exercise in Tallinn brings 4000+ experts to simulate large-scale cyberattacks and strengthen defense readiness.
🎙️ And I am extremely excited to announce that I launched my own very personal and very much-awaited podcast: Inside Finance with Gerlyn Tiigemäe.
The first episode is already live, and I could not have asked for a better guest to start with. I sat down with Mari-Liis Kikas, a fractional CFO who has worked across more than 10 companies and brings a very honest view into what finance teams are really struggling with when it comes to AI adoption.
🎧 Listen to Inside Finance 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, next trainings for beginners @ 14.05 – information; for intermediate level @ 05.06 – information
- Training “AI võimalused projektijuhtidele” @ Äripäeva Akadeemia, next trainings @ 15.06 – information
