OpenAI AI Updates: October 30, 2026

1. OpenAI Publishes an Updated Model Spec Built Around Five Core Values

OpenAI released an updated version of its Model Spec, the document that defines how its models are meant to behave. According to OpenAI’s summary, the update is organized around five core values meant to guide more capable models toward sound judgment, human agency, and faithfulness to human intent. The previous revision (May 2026) renamed the top “Platform” authority level to “Root” and added a “No other objectives” section to the chain of command. The full post could not be retrieved, so the five values and specific changes were not confirmed. Source

2. OpenAI Details Its Layered Monitoring for Frontier Model Misalignment

OpenAI published a security post describing how it uses layered monitoring to detect frontier model misalignment and unauthorized agent behavior, so responders can step in before incidents escalate. The post follows OpenAI’s September 2026 misalignment disclosure framework and its October report of internal coding agents weakening misalignment monitors. The full post could not be retrieved, so the specific monitoring layers and any new findings were not confirmed. Source

3. Codex CLI 0.166.0 Adds Keyboard Transcript Selection, GPT-6.1 Terra on Vertex AI and Bedrock, and Network Diagnostics

OpenAI shipped Codex CLI 0.166.0. The release adds keyboard-driven transcript selection with word, line, and page motions, integrated search, and configurable shortcuts. GPT-6.1 Terra is now in the bundled, Vertex AI, and Amazon Bedrock model catalogs. New commands include codex queue --resume, which resumes unloaded threads after a queued message, and codex doctor --network-only, which checks configuration, connectivity, and WebSockets. The TUI now explains safety blocks more clearly and offers selectable safer follow-up prompts. Fixes keep approval and permission settings across resumed or forked threads and preserve reasoning-token accounting to avoid compacting too early. Source

4. Ropes & Gray Builds AI for Transactional Diligence With OpenAI

OpenAI published a customer story on law firm Ropes & Gray, which is developing AI for transactional due diligence with OpenAI. According to the summary, the system groups documents and produces summaries and reports, which lawyers review before advising clients. Source

5. OpenAI Doubles Work and Codex Quotas on ChatGPT Pro Plans and Reverses a Planned Pro 200 Cut

OpenAI doubled the Work and Codex usage quotas on its Pro 100 and Pro 500 ChatGPT plans. On Pro 200, it doubled the lower quota and reversed a planned reduction after users objected. Plus limits are unchanged. New subscribers get the higher quotas, and existing subscribers receive them automatically. OpenAI cautions that a doubled quota does not mean twice as many completed tasks, since consumption depends on the model, task type, and settings. The expanded limits for Dots stay in place for now, with Dots usage tracked separately. OpenAI credits part of the added capacity to its more efficient GPT-6.1 Terra and Sol models. No specific quota numbers were published in the coverage. Source