Transcript · 2026-10-29

Welcome to Overnight AI. I'm Alex. And I'm Jamie. We are your AI-generated hosts, and here's your AI news briefing for October 29th, 2026. Today, enterprise workflows, limited access to a new model, cloud earnings, and early university research. OpenAI and ServiceNow are expanding their partnership to connect ServiceNow workflows with ChatGPT and Codex. Customers would bring context from their own instance into ChatGPT and expose selected actions, such as investigating an incident, assessing a change, or updating a record. What keeps that from becoming blanket access to a company's systems? OpenAI says ServiceNow's Action Fabric will apply existing business rules and permissions. Customers choose the capabilities they expose. The intended benefits, therefore, depend on each organization's configuration, not just on the model understanding a request. And availability is mixed. The ServiceNow MCP plug-in is available now, but enhanced capabilities are expected by the end of 2026. The planned security offering is slated for early access in January 2027 and general availability in the second quarter. ServiceNow also plans to expand ChatGPT work and Codex beyond engineering to all employees. Those are product plans and intended workflow benefits, not independently measured deployment outcomes. Connecting a model to business software still requires permissions, review, and accountability for the actions it takes. Controlled access also figures in Google's model news. In Alphabet's quarterly remarks, Google said it had announced Gemini 4 Carbon, which it describes as its strongest frontier model for cyber security and long-horizon work. Google says Carbon will first reach a subset of trusted cyber defenders and is being tested by the United States government before wider access. It says the model leads the updated Artificial Analysis Intelligence Index and outperforms Gemini 4 Argon across several areas. Does that mean developers can use it broadly now? No. Google's remarks did not announce general availability. Its performance descriptions are company claims, including its account of the external index results. The concrete detail is a staged rollout. Limited access does not itself prove safety, but it does distinguish this announcement from a broad consumer or developer release. Alphabet also reported third-quarter revenue growth of 25% year-over-year and Google Cloud growth of 103%. It said first-party model APIs were processing nearly 32 billion tokens per minute, up 40% from the previous quarter, and the Gemini app had reached 1 billion users. Google says Gemini 4 Argon's first-week developer platform usage ramped 3 and 1/2 times faster than Gemini 3, despite capacity constraints. Those are company-reported measures. They describe demand and infrastructure pressure. They do not independently establish model quality or a particular customer's results. Microsoft offered another view of cloud demand. For the quarter ended September 30th, it reported revenue of $92.3 billion, up 19%. Microsoft Cloud revenue reached $63.9 billion, up 30%, while Azure revenue increased 48%. It also reported commercial remaining performance obligations of $677 billion, up 73%. That measures contracted future work, not revenue already booked. Microsoft's release does not isolate the contribution from any one AI product, so we cannot assign all that cloud growth to AI. The company separately identified the effect of its OpenAI investment in its adjusted accounting. GAAP net income was $36.6 billion, compared with adjusted non-GAAP net income of 36.4 billion. The distinction matters when comparing the headline figures. Our final story is about research funding. Meta says it is providing $2 million in grants and model access across five US university partnerships: Caltech, Ohio State, Louisiana State, Indiana University, and UT Health Houston. The projects span medicine and several scientific fields. Some medical examples remain explicitly early. Louisiana State's eyes-on project is developing computer vision to detect possible tracheostomy tube displacement and distress in infants. Meta says a proof of concept has been published with more work needed. Ohio State researchers are investigating prediction of future heart failure risk using imaging, records, and ECG data when available. Those are research projects, not clinical validation or evidence of improved patient outcomes. The grants and access arrangement are the news. Health claims still need results and appropriate evaluation. That same discipline runs through today's briefing. Distinguish availability from plans, demand from outcomes, and research from established care. For the original announcements, results, and qualifications, find the source links in the show notes. Thanks for listening to Overnight AI.