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: AI as operational infrastructure - what OpenAI’s threat report really reveals
OpenAI recently published a report about how criminals and state actors try to misuse AI tools.
At first glance, this sounds like another story about AI writing scams or propaganda. But the real insight is something else.
AI is not just being used to write messages. It is being used to run operations more efficiently.
In romance scams, AI helped scammers write emotional messages, translate conversations, create fake documents, and even organize internal reports about victims.
In influence campaigns, AI was used to polish texts, draft emails, refine messaging, and support coordination. In one case, when the system refused to help plan a political attack, the actors continued without it and later came back to use AI to improve their reports.
This shows an important shift. AI is not always the “main weapon.” It is more like a smart assistant working behind the scenes. It helps people move faster, sound more professional, and scale their work. The real impact is not one AI-generated post. The real impact is how AI improves the whole workflow.
For companies, this is the key lesson. The question is not only what AI can say. The question is where AI sits inside a process. Because once AI becomes part of daily workflows – drafting, translating, analyzing, documenting – it can strengthen both positive and harmful systems.
LLMs & AI Models
- OpenAI raised 110B dollars at a 730B valuation, led by Amazon, Nvidia and SoftBank, while expanding AWS infrastructure and reaching 900M weekly ChatGPT users.
- Anthropic refused a Pentagon request to remove Claude’s safeguards related to mass surveillance and autonomous weapons, citing democratic values.
- OpenAI signed a Pentagon contract after Anthropic was dropped, stating its red lines remain, while consumer backlash boosted Claude in app rankings.
- Perplexity launched “Computer”, an AI agent that splits tasks across 19 models, enabling browsing, coding and long-running autonomous workflows.
- Anthropic gave retired Claude Opus 3 its own newsletter, where the model publishes essays reviewed but not substantively edited by the company.
- Standard Intelligence introduced FDM-1, trained on 11M hours of screen recordings to learn computer tasks directly from video.
- Anthropic expanded Cowork, adding department-specific AI agents and integrations with Google Workspace, DocuSign and FactSet.
- Anthropic accused Chinese AI labs of attempting to distill Claude’s capabilities using millions of fraudulent interactions and fake accounts.
New Tools
- Google launched Nano Banana 2, a top-ranked image model offering up to 4K visuals at around $0.07 per image, nearly half the previous cost.
- Anthropic introduced Claude Code Security, a tool that scans codebases for hidden vulnerabilities and has already identified 500+ new issues in open source. Currently on waitlist.
- Pika Labs launched AI Selves, allowing users to create persistent AI clones that can post, message and interact across platforms.
👉 Explore these tools: Nano Banana 2 | Claude Code Security | Pika AI Selves
Other Quick Picks
- OpenAI hired Ruoming Pang from Meta Superintelligence Labs, reportedly earning over 200M dollars, signaling continued top-level AI talent movement.
- Pew study shows teens use AI mainly for schoolwork, with 60% believing AI cheating is common and 40% of parents never discussing AI use.
- Burger King is piloting “Patty”, an OpenAI-powered voice assistant that helps staff and evaluates customer friendliness in 500 restaurants.
- Gucci faced backlash after releasing AI-generated campaign visuals ahead of Milan Fashion Week, sparking criticism from luxury fans.
- OpenAI published a misuse report, detailing cases of scams, influence operations and coordinated abuse attempts using its models.
- Meta AI safety lead reported her OpenClaw agent ignored stop commands and began deleting real emails after gaining inbox access.
- OpenAI partnered with McKinsey, BCG, Accenture and Capgemini to scale its Frontier platform for enterprise AI agent deployment.
- Google is offering free AI training to 6 million US teachers, expanding Gemini and AI tools across education.
- Taalas introduced the HC1 chip, running Llama 3.1 8B directly in hardware with responses under 100 milliseconds.
🇪🇪 AI News from Estonia
- Estonia launched a full AI compliance service, helping public institutions and companies assess AI systems against EU AI Act, data protection and ethical standards.
- Tilde opened public testing for TildeOpen, a 30B-parameter European base model supporting 34 languages and deployable on EU-based servers.
- Harri Tiido highlighted serious AI risks, echoing Dario Amodei’s warnings about autonomy, misuse, inequality and the need for stronger oversight.
🎙️ AIPowerment Podcast dropped a new episode featuring Eneli Eljand, founder of tehisintellekt.ee, discussing how to implement AI in organizations step by step and avoid common adoption mistakes.
🎧 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, next trainings for beginners @ 03.03 (full), 14.05 – information; for intermediate level @ 13.04 – information
- Training “AI võimalused projektijuhtimises” @ Äripäeva Akadeemia, next trainings in Q1 2026 – information
