OpenAI AI Updates: November 2, 2026

1. OpenAI SDKs Add Context Compaction, Time Awareness, and Session Review Settings for Beta Agents

OpenAI released openai-python v3.42.0 and openai-node v7.46.0 on October 31, exposing several new API surfaces. Beta agents, sessions, and scheduled agent runs gain context compaction settings that choose between background and blocking compaction (the async_ parameter in Python), an optional experimental.time_awareness setting whose updates merge by key, and session review settings plus the ability to replace a session’s tools through agent.tools. Image edits accept a transparent PNG markup overlay to guide the edit; it must match the first image’s dimensions and cannot be combined with a mask. The admin API adds organization external key registration and validation for AWS, Google Cloud, and Azure, SIP sideband connections can send keypresses with transport.dtmf.send and receive transport.dtmf.sent acknowledgements, and the Node release also exposes hosted tool usage in responses and filters agent sessions by metadata. Follow-up releases (Node v7.47.0 on October 31, Python v3.42.1 on November 1) add typed Responses WebSocket warmup and fix Responses WebSocket connection behavior. Source

2. ChatGPT Plus, Pro, and Business Plans Will Share One Token Budget Across ChatGPT and Codex Starting in December

OpenAI updated its help center page on ChatGPT and Codex usage limits ahead of merging Chat and Work, which folds Work’s agentic features into the main ChatGPT interface. According to The Decoder, starting in December, Plus, Pro, and Business plans will draw from a single token budget shared across ChatGPT and Codex, with the remaining balance and reset times shown in Settings. Chats with Instant models do not count toward the budget, and users who hit the limit can wait for a reset, upgrade, or buy extra credits. ChatGPT currently asks users to choose between two operating modes and 11 model settings, and OpenAI doubled usage limits on its higher-priced plans earlier in the week. Source

3. ChatGPT Adds Card and Device Limits on Paid Subscriptions

OpenAI introduced card and device limits for paid ChatGPT accounts, describing them as a security measure in a new help center article. Users who link too many subscriptions to one payment card cannot start new ones, and each device is capped on the number of paid accounts it can use; once the cap is hit, ChatGPT shows “Device limit reached,” the affected account can no longer be used on that device, and logging out does not free a slot. The limits apply to Go, Plus, Pro 100, Pro 200, and Pro 500, while Free, Business, Enterprise, and Edu plans are excluded, and OpenAI has not published the exact caps. The Decoder notes the change mainly affects users who rotate several cheaper subscriptions on one device for more usage; OpenAI points them to extra credits or usage resets instead. Source

4. OpenAI’s Chris Lehane Outlines a Post-Midterm Push, Including Calling Data Centers “AI Compute Campuses”

OpenAI is preparing a public affairs push for after the November 3 U.S. midterm elections, chief global affairs officer Chris Lehane told Axios, as reported by TechCrunch. Lehane called the midterms “a bit of a movie trailer for what you’re going to see in 2028” and said he expects AI to become “the issue” in the presidential race. Ideas he outlined include AI “speed limits,” helping people see AI as empowering in daily life, pushing the U.S. to help develop global AI standards, and rebranding data centers as “AI compute campuses.” Lehane called the public debate over data centers “good and healthy” and said communities should hold companies accountable for delivering real benefits from these projects. Source