Hello AI Enthusiast,
This week the industry spent a lot of words on slowing down and a lot of code on speeding up. Anthropic published a three step plan to pace the frontier and most of the other labs nodded along, while Nvidia's CEO said safety should be left to the engineers who build it. Then the same companies shipped something much more specific: ChatGPT and Claude both went after finance in opposite ways, and OpenAI and Google released voice models good enough to answer your phone. Let's get into it.
The Big Picture 🔊
Anthropic wants to slow AI down, and everyone is talking
Dario Amodei published We Must Pace the Frontier, a three-step plan to slow the rate of AI capability gains. It starts with third-party evaluators embedded inside frontier labs, something Anthropic is committing to now, then coordination between democratic countries, then global talks. TechCrunch reports OpenAI, Anthropic and Google DeepMind have been discussing safety for weeks. Nvidia's Jensen Huang disagrees, and Microsoft published its own Humanist AI Code of Conduct.
|
||
|
Many major players are calling to slow down at once, and a real framework would need labs and governments to agree, so how far it goes is still open. For companies outside the frontier labs, this debate changes little day to day. The tools available now keep working while the bigger discussion plays out. |
OpenAI and Claude go after Finance, in opposite ways
Three days apart, both labs shipped for finance professionals. ChatGPT for Financial Services puts premium market data inside ChatGPT, runs on GPT-6 Astra, and goes through their sales team. Claude for Financial Advisors takes the other route, with connectors and skills that plug into the data tools advisors already pay for, available to Claude Enterprise customers.
|
||
|
Claude hands you the parts and assumes you already have the subscriptions. OpenAI hosts the data and asks you to talk to sales. One costs you setup time, the other probably costs you more money. |
Voice AI is now good enough to interrupt you
OpenAI released GPT-Live-1 in the API, one model that both listens and speaks, so it can be interrupted mid-sentence and answer back naturally. Google replied with Gemini 3.8 Live and Live Extended Thinking, which takes turns like a walkie-talkie but can run tasks in the background while it talks. Yelp is already using GPT-Live to answer restaurant and plumber calls.
|
||
|
Two approaches again. OpenAI went for natural interruption, Google for turn taking with work happening in the background. We will fully believe the demos when an AI successfully books a plumber in Italy. Until then, keep in mind that some people think out loud and some people need to write things down. Voice will not replace typing for everyone, and offices are about to get much louder. |
The finance tools launched this week assume your team already knows how to work with AI. Most teams do not, and the distance between the two keeps growing. Our corporate training helps your people find where AI actually saves time in their own work, and how to use these tools without starting from zero.
Bits and Bobs 🗞️
OpenAI's Agents API lets developers build and run agents with managed infrastructure and custom compute environments in a single API call.
OpenAI's Data agent connects to business data sources so you can ask questions and build dashboards in plain language.
Claude for Small Business added workflows, integrations and training programs for teams without an IT department.
Microsoft Copilot Studio now lets people build enterprise apps by describing them, with admin controls attached.
Anthropic's September threat report details how state-backed groups and criminal networks used Claude for cyberattacks, surveillance and influence operations.
Sam Altman said OpenAI will not go public this year, calling it an ill-advised moment and pointing to safety work.
Anthropic and Accenture published a blueprint for moving enterprise AI from pilot projects into production.
ChatGPT now has pets, little companions that track your chats and can start new ones.
From Our Founder’s Channels 🤳
In his latest LinkedIn post, Gianluca Mauro shares the case study he presented at the Harvard Chan Advanced Learning Academy on the lessons learned from building Epiphany AI. Read it here.
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! 👋
