Uber is rolling ChatGPT Work from a 1,000-employee pilot to more than 30,000 people company-wide, with GPT-6.1 Sol as the default.1 AWS opened SageMaker training sessions, a serverless custom-training loop the customer writes while SageMaker handles the GPUs.2 OpenAI pushed ChatGPT Ads into 26 more countries, 89 in total, and says more than 100,000 advertisers have used the platform.3 Anthropic invited principal investigators in the life sciences to eight-to-twelve week collaborations using Claude.4

In the news

Uber rolls ChatGPT Work to 30,000 employees

OpenAI posted on Oct 22 that Uber is rolling ChatGPT Work, an agent inside ChatGPT Enterprise for multi-step knowledge work, from a 1,000-employee pilot to more than 30,000 employees. Uber CTO Praveen Neppalli Naga and OpenAI president Greg Brockman are quoted on the page. The companies call it Uber's first company-wide deployment of an agentic platform for knowledge work.1

ChatGPT Work connects Google Drive, Gmail, Google Calendar, and Slack. GPT-6.1 Sol is the default model. Voice and image generation sit alongside it.5 6

Uber will back the rollout with an internal AI skills library, approved integrations, and training across operations, finance, marketing, and product. Developers keep using Codex. The companies say they will share adoption, use cases, and measurable outcomes as the rollout proceeds. ChatGPT Work shipped on Jul 9, and GPT-6.1 Sol is the default inside that product.

SageMaker training sessions generally available

AWS said on Oct 22 that training sessions are generally available on SageMaker AI. A team writes the post-training loop, including custom loss or reward signals. SageMaker allocates GPUs, runs the distributed job, recovers from faults, and stores checkpoints. The customer does not pick instance types or write distributed-training config.2

Named techniques include SFT, DPO, RLVR, RLAIF, and multi-turn RL, with LoRA or full fine-tuning and continuous customization. A cookbook and agent skills run in SageMaker Studio, Kiro, Claude Code, Codex, or GitHub Copilot. Billing is by tokens processed, and the meter pauses while a reward function scores a batch in the customer's environment. Data and checkpoints stay in the customer's S3, under IAM, with optional KMS and VPC.7 8

Named models at launch are gpt-oss-20b, gemma-4-31B-it, Qwen3.6-27B, and Qwen3.8-27B. AWS quotes ImagineSoftware's Rob Gontarek on healthcare claim-denial triage. Karini AI's Nitin Wagh is quoted on gpt-oss-20b for multi-turn tool calling against Amazon Neptune.2

ChatGPT Ads expand to 89 countries

On Oct 21 OpenAI said ChatGPT Ads now reach 26 more countries, 89 in total. OpenAI reports more than 1.2 billion weekly ChatGPT users and more than 100,000 advertisers since ads launched just over eight months ago. Ads stay on Free and Go. Plus, Pro, and Enterprise remain ad-free.3

New markets include 14 more Latin American countries (Argentina, Chile, Colombia, and Peru among them), plus Nigeria, South Africa, Serbia, Ukraine, Cambodia, Bangladesh, and Pakistan. Advertisers get in through technology partners or self-service Ads Manager at ads.openai.com. OpenAI says conversations stay private from advertisers, ads are labeled and separate from answers, and advertising does not influence ChatGPT's answers. L'Oréal Group CDMO Asmita Dubey is quoted. OpenAI names more formats, objectives, buying options, measurement, and more native business interactions as next work.9 10

Anthropic guest investigators in the life sciences

Anthropic opened an Oct 22 call for guest investigators in the life sciences. Principal investigators and research groups at academic institutions, nonprofits, or government labs can propose eight-to-twelve week collaborations using Claude on causal human biology, structural and computational molecular biology, medicinal chemistry, drug discovery, and clinical science. Apply at anthropic.com/guest-investigator-program.4 11

Anthropic researchers run models and harnesses at no cost to the guest. Outputs (analyses, artifacts, tools) are to be published with open-sourced data and code. Guests keep rights to their data and to independent follow-on research. Anthropic is not asking applicants to hand over data. Applications should use only publicly available information.

Applications stay open about 14 days. Selected participants are notified about 14 days after close, with an extension possible. Anthropic does not say how many slots are open. Sibling programs named on the same post are the Life Sciences Verification Program and AI for Science credits. This is a research collaboration call, not a new Claude model.12