AI’s Next Move: Agents, Safety Audits, and Big Partnerships

Anthropic launches beta reflection dashboard for Claude users Claude usage dashboard shows your chat patterns, topics, and habits over time Review AI use, set quiet hours, and see where human judgment still matters ChatGPT Work adds agentic features for tasks across apps and devices ChatGPT Work streamlines multistep projects with GPT5.6 and Codex Get documents, slides, and apps faster across web, mobile, and desktop GPT5.6 becomes the preferred model in Microsoft 365 Copilot GPT5.6 in Microsoft 365 Copilot boosts writing, analysis, and presentation work Use it across Word, Excel, PowerPoint, and Chat for faster, smarter results AWS GraphRAG deployment speeds drug research cycles AWS GraphRAG cuts drug R&D cycles 87% with one naturallanguage knowledge graph Unify clinical, lab, and research data to speed discovery and keep context intact OpenAI audits coding benchmark and finds flaws SWEbench Pro audit shows 30% of tasks are broken, skewing coding results. See why OpenAI withdrew its recommendation and what it means for model evals. OpenAI expands bio bug bounty program OpenAI Bio Bounty Program targets universal jailbreaks in frontier models Top awards reach $50K; apply to test GPT5.6 and future versions OpenAI publishes national security partnership principles OpenAI National Security Principles guide safe AI use in government and defense See how the company limits surveillance and autonomy while backing safeguards Anthropic invites public questions about AI Anthropic asks hard AI questions to surface public concerns on jobs and society See how it plans to answer them and track its publicbenefit mission Ben Bernanke joins Anthropic’s LongTerm Benefit Trust Ben Bernanke joins Anthropic's trust to guide AI governance and longterm impact. See how the exFed chair could shape decisions on markets, work, and society.