Hello AI Enthusiast,
This week, the open weight debate turned political, with most of the industry signing a letter against restrictions while Anthropic stood apart. Meanwhile, OpenAI and Anthropic both pushed voice mode forward, and Microsoft struck a multibillion dollar deal with Mistral to bring more AI compute to Europe. Let's get into it.
The Big Picture 🔊
Tech giants sign open letter defending open weight AI
Nvidia, Microsoft, Meta, Palantir, and over 20 other companies signed a letter urging policymakers to avoid restricting open-weight AI models, warning this could push innovation overseas. Google and OpenAI later joined. Anthropic and Amazon did not sign. Anthropic CEO Dario Amodei responded publicly, denying he supports banning open-weight models but arguing China-related risks need targeted measures like chip restrictions and limits on distillation.
|
||
|
Anthropic wants safety testing for capable models, even open ones, but that's hard to enforce once weights are downloaded and modified freely. The letter's real target seems to be Chinese competition, not open weights itself. Cheaper open models, like DeepSeek, are also becoming a business threat to closed-model providers. |
OpenAI and Anthropic both roll out voice mode upgrades
OpenAI launched GPT-Live, a full-duplex voice model that lets ChatGPT listen and speak at once, replacing Advanced Voice Mode. It also brought voice control to the desktop app, letting users direct agents in ChatGPT Work and Codex. Anthropic upgraded Claude's voice mode too, adding Opus and Sonnet model options beyond Haiku, plus the ability to act on connected apps like Gmail, Calendar, Slack, Notion, and Canva mid-conversation, with permission required before acting.
|
||
|
The experience is different too. Voice responds with pauses, small laughs, and shorter answers, likely designed to feel more human and pleasant to talk to. Running voice mode also costs more than text, since audio tokens are priced higher. Still, the whole industry seems to be betting on voice as the next way we'll work. |
AI features like voice mode keep piling up, but most companies are still figuring out how to actually put them to work. Our Corporate AI Training helps teams cut through the noise and focus on what works for them.
Microsoft signs multibillion dollar deal with Mistral
Microsoft and Mistral expanded their partnership with a multibillion dollar deal, letting Azure customers use Mistral's data centers in France for extra compute in Europe. Mistral's Medium 3.5 and OCR 4 models joined Microsoft Foundry and Copilot Studio, and businesses running Azure Local can now deploy Mistral's open models with full control over data and infrastructure. The deal adds no new equity stake.
|
||
|
Microsoft was the one big player without its own frontier model, so partnering with Mistral fills that gap while keeping data inside Europe. Running models and compute on Mistral's French data centers makes compliance easier for regulated industries, and could push more European companies toward adopting Microsoft's AI stack. |
Bits and Bobs 🗞️
Anthropic released Claude Opus 5, a more capable and efficient model that improves on coding, scientific research, and visual tasks, and published a guide to its model families (Mythos, Opus, Fable, and Haiku) to help you pick the right one for your budget, speed, and complexity needs.
Anthropic's Claude Code AI now uses improved judgment and progressive disclosure to handle code review and tool usage without strict system prompt rules or repetitive instructions.
Microsoft launched Perception, an AI cybersecurity platform with the MAI-Cyber-1-Flash model, claiming superior performance in detecting and fixing cyber threats using AI-driven red, blue, and green teams.
OpenAI Presence is a new enterprise product that deploys trusted AI agents with built-in policies and guardrails to reliably handle high-value tasks and adapt over time with human oversight.
OpenAI published a report on how AI is letting workers take on tasks beyond their usual roles, reshaping job responsibilities and the division of labor across occupations.
ChatGPT Health lets U.S. users securely connect Apple Health and medical records so AI can provide personalized health insights and track changes over time while ensuring privacy.
The European Commission fined Google €890 million for favoring its own services under the Digital Markets Act, and separately ordered it to give third-party AI assistants equal access to Android and share anonymized search data with rivals.
The latest MCP release introduces a stateless request/response model and enhanced security, making it easier to build and scale AI applications with Claude.
From Our Founder’s Channels 🤳
In his latest LinkedIn post, Gianluca Mauro takes readers behind the scenes of Epiphany's agentic course creation and the research that shaped it. Read it here.
LOLgorithms 😂
If voice mode is the direction we're going, we'll witness a lot of interesting conversations in the street.
That's a wrap on our newsletter!
If you're looking to build AI skills in your team, check out our Corporate Training. Every program is tailored to your team and built around hands-on practice.
Catch you next week! 👋
