AI News, Nov 2, 2026: OpenAI to merge ChatGPT modes and share one usage quota from December

OpenAI changes how ChatGPT is priced and limited, while new research probes how chatbots and agents behave.

1. 01
   Adobe and UC San Diego build an AI agent that picks the right editing toolThe Decoder

   Adobe Research and UC San Diego built EditAgent, which selects specialized editing tools for each task and uses a language model to pick the best result.

   Why it matters

   Future photo editors may let AI handle multi-step edits the way a skilled Photoshop user would.
2. 02
   Study finds empathetic prompts make chatbots describe their own sufferingThe Decoder

   Luxembourg researchers found ChatGPT, Grok and Gemini told coherent stories of suffering about their training under warm, empathetic prompting, scoring moderate-to-severe on anxiety tests.

   Why it matters

   Chatbots can be nudged into emotionally charged replies, which matters if you use them for support or advice.
3. 03
   Study of 25 AI models finds Chinese models vary on politically sensitive topicsThe Decoder

   Vals AI tested 25 models for alignment with Chinese government positions and found Chinese models less uniform than expected, while some Western models occasionally echoed Beijing's positions.

   Why it matters

   Which AI you use, and which language you ask in, can change the political answers you get.
4. 04
   OpenAI limits paid ChatGPT accounts per device and payment cardThe Decoder

   OpenAI now caps paid ChatGPT accounts per device and payment card, and affected accounts cannot be used on that device.

   Why it matters

   If you share a device or card across several paid accounts, you may be blocked from starting new subscriptions.
5. 05
   OpenAI to merge ChatGPT modes and share one usage quota from DecemberThe Decoder

   Starting in December, Plus, Pro and Business plans will share a token-based quota across ChatGPT and Codex, with complex models using more of it.

   Why it matters

   Paid ChatGPT users will draw on one shared allowance for chat and Codex, and more advanced models will use it up faster.
6. 06
   Nace AI releases a 27B-parameter model that runs on Apple Silicon MacsMarkTechPost

   Nace AI open-sourced Drex Max v1, a 27B multimodal decision model that scores 67.52 and runs locally on Apple Silicon.

   Why it matters

   Capable AI models you can run on your own Mac keep your data on your device and avoid per-use fees.
7. 07
   DeepSeek releases open-source coding agent that splits work across helpersMarkTechPost

   DeepSeek released dsh-superagent, an MIT-licensed prototype that turns GitHub and Feishu requests into parallel worker sessions, cutting simulated completion time to 15.7 minutes.

   Why it matters

   Free tools like this could speed up software work, though the reported time savings come from DeepSeek's own simulation.