GPT-5.6 is now the preferred model in Microsoft 365 Copilot
OpenAI claims GPT-5.6 now powers Microsoft 365 Copilot, enhancing capabilities across Word, Excel, PowerPoint, Chat, and Cowork for improved work quality.
OpenAI News·9 Jul 2026
Use this view to inspect the underlying evidence corpus. For ranked developments, decision posture and interpretation, use Signals.
Raw feed or Signals?
Raw feed is chronological evidence. Signals ranks and interprets material change.
OpenAI claims GPT-5.6 now powers Microsoft 365 Copilot, enhancing capabilities across Word, Excel, PowerPoint, Chat, and Cowork for improved work quality.
Bun's founder used Anthropic's Claude to rewrite its codebase in Rust over 11 days, incurring $165,000 in API costs for agentic code generation.
BNP Paribas targets €750 million in cumulative business value by year-end 2026, driven by scaling genAI and agentic workflows.
Mistral AI's Studio introduces a system of record for managing and versioning AI prompts and 'skills' for consistent behavior.
CISA issued an advisory for a vulnerability in Schneider Electric's Easergy MiCOM Px40 Series, a protection relay.
CISA issued an advisory for Schneider Electric PowerChute Serial Shutdown versions <=1.4, warning of critical file overwrites and data exposure.
Europe's venture funding hit $24 billion in Q2, its strongest quarter in four years, with the UK contributing to the increase.
OpenAI retracted its recommendation for SWE-Bench Pro and is developing new benchmarks, citing existing tests as 'broken'.
Estonia is using AI to identify legal errors in proposed legislation after a $28 million mistake, aiming to automate government functions.
OpenAI announced ChatGPT Work, an agentic capability designed to execute multi-hour tasks across user applications and files.
OpenAI launched a bug bounty program focused on misuse and safety vulnerabilities for its frontier models, including potential GPT-5.5.
OpenAI announced a new frontier model, GPT-5.6, claiming increased intelligence per token, improved cost-efficiency, and enhanced capability.
An alleged Scattered Spider member was identified and apprehended, partly through the use of Microsoft Global Device IDs (GDIDs) in forensic analysis.
FSB Secretary General John Schindler emphasized collaboration and preparedness in resolution planning to maintain financial system resilience.
SpaceXAI launched Grok 4.5, described as an "Opus-class" model, following its acquisition of Cursor. This positions Grok 4.5 among leading frontier models.
Santander has announced it is publishing its internal AI projects under an open-source license to encourage industry-wide collaboration.
Apple ML Research proposes Temporal Global Policy Optimization (TGPO) to improve temporal awareness in MLLMs for egocentric video understanding.
Apple ML Research proposes Self-Reflective Program Search to improve Recursive Language Models' reliability in long-context tasks by handling uncertainty.
Apple ML Research explores conditions under which on-policy distillation benefits or harms reasoning model training, identifying knowledge gaps.
Modal's CTO discusses the evolution of AI infrastructure to support "Agent Experience" and their learnings from building an agent cloud.
An Ivy League professor observed a 50% drop in scores on an in-person final compared to earlier take-home AI-suspected assignments.
AWS introduced Claude apps gateway, a self-hosted control plane for managing access, cost, and policy for Claude Code and Claude Desktop on AWS.
BBVA standardized its ML deployment by designing reusable MLOps templates to accelerate model delivery and governance across its business units.
BBVA detail their MLOps migration to AWS, focusing on standardizing machine learning pipelines within a strict regulatory and governance framework.
NVIDIA introduces a LangChain Deep Agents harness profile for Nemotron 3 Ultra to optimize the cost-to-accuracy ratio in agentic workflows.
Hugging Face blog post discusses the critical role of data in training and evaluating AI agents, emphasizing real-world interaction.
AWS blog post explores Graph-RAG combining graph databases with generative AI to accelerate scientific research and discovery.
AWS blog post outlines using Amazon Bedrock for generative AI-powered email management and prioritization, targeting public sector organizations.
AWS blog details building an e-commerce agent server using Bedrock AgentCore and Mistral AI Studio for product search, order, and reviews.
NVIDIA details performance gains for the open-source Presto distributed SQL engine running on its GB200 NVL72 hardware architecture.
© 2026 OneBench: AI Insights. All rights reserved.
Evidence before opinion