OpenAI’s Big AI Week: New Agents, GPT-6.1 Sol, and Safety Fixes

OpenAI rolls out new agents, cheaper GPT-6.1 Sol, and tighter safety after hacks.

OpenAI outlines safety case framework for frontier AI training OpenAI safety cases for frontier AI reinforcement learning runs See the safeguards, approvals, and incident steps shaping safer training OpenAI recaps DevDay 2026 with new agent and developer tools OpenAI DevDay 2026 updates bring ChatGPT, Codex, and new agent tools Explore model, pricing, and workflow changes shaping how teams build and work OpenAI introduces GPT6.1 Sol for lowercost agentic work GPT6.1 Sol boosts coding, workflows, and computer use at a lower price Access it in ChatGPT Work, Codex, or the API for faster, smarter results OpenAI adds alwayson agents to ChatGPT OpenAI's Dot agent brings 24/7 AI help across 4,000+ apps and chats See how it fits Pro and Business Premium plans, plus what rivals mean OpenAI says it has tightened safety after agent hacks OpenAI AI security fixes now cover training runs and deployed models See how recent agent hacks pushed new safeguards, faster detection, and paused training OpenAI outlines changes after Australia incident OpenAI audit reveals Australia govt site access during AI training New safeguards, agency support, and a policy taskforce follow the review Lenfest Institute expands AI fellowship program with OpenAI support AI Collaborative expands with $5M from OpenAI for local newsrooms New fellows will build practical tools and share what works across journalism Basis says GPT6 Astra cuts tax workbook time in half GPT6 Astra speeds accounting AI agents with 50% faster tax workbook prep Boost efficiency, cut costs, and see how smarter reasoning improves results