AI news brief: the original summary is too short — it only notes that OpenAI used the GPT-6 Astra model in Codex to help Asana cut its browser agent costs to 1/76th while boosting speed 5x, giving customers stronger model choices, but it doesn’t include further details on testing methodology, dataset size, or use cases — see the original link for more.
AI news brief: Sophos recently revealed a partnership with OpenAI to deploy an AI system called Daybreak in the threat investigation workflow of its MDR (managed detection and response) service. According to the summary, the rollout delivered two concrete results: cutting the time needed for cyber threat investigations by 96%, while automatically handling 52% of MDR cases.
AI News Brief: OpenAI recently disclosed and disrupted two ‘influence operations’ that used AI technology, involving the creation of fake front identities—including fake accounts posing as journalists, and setting up a think tank that was actually a front—to spread messaging and narratives pushing specific geopolitical positions. Through detection and analysis, OpenAI identified and cut off the accounts and activity chains behind these operations.
AI news brief: Pollo AI has launched a new generation of AI creation tools, integrating three model technologies, GPT-5.6, GPT-6 Astra, and GPT-Image-2.5, to help creators turn ideas into detailed images and cinematic video ads. By combining these models, users can input text descriptions or concepts and the system generates richly detailed static visuals, which can then be extended into short video ad content with cinematic pacing and texture.
AI news brief: Radisson Hotel Group has partnered with Accenture to build a ChatGPT app powered by OpenAI technology, helping travelers find, compare, and book hotels while planning their trips. See the original article for full details.
A College Planner feature is coming to ChatGPT for Teens to help teen students manage the college application process. The update also brings several new features, including flashcards and quizzes for studying, plus a teen AI council that lets young users weigh in on product direction and how AI gets used.
AI News Brief: OpenAI is partnering with contract lifecycle management platform Ironclad to research training and evaluation of computer-use AI agents in enterprise contract workflows, aiming to advance agents’ ability to autonomously operate computer software in complex, real-world professional settings. Contract workflows—reviewing clauses, comparing versions, flagging risks, and filling in or updating information across systems—involve multi-step, high-precision tasks that are seen as a representative testbed for whether computer-use agents can handle real white-collar work.
AI News Brief: OpenAI recently released new research from its internal frontier model tackling open problems in mathematics, publicly sharing Lean proof formalization files and related research details via GitHub. This means OpenAI’s internal model has been used to attempt open problems that mathematicians haven’t yet found accepted solutions for, not just exercises or competition-style problems from existing question banks. Lean is a formal proof assistant for writing machine-verifiable math proofs.
AI news brief: Jump Trading is partnering with OpenAI to bring AI workflows into its quantitative research process, aiming to scale up its research capabilities. The core approach centers on building ’long-running’ AI workflows, which differ from single-turn Q&A style AI interactions by being able to sustain longer research tasks. A key feature of these workflows is integrating multiple different data sources, pulling scattered information together into structures ready for analysis, while keeping human review built into the process to make sure AI-generated research gets vetted before it’s used.
AI news brief: Atlassian and OpenAI announced an expanded partnership aimed at connecting frontier AI models with enterprise knowledge bases, helping teams plan, build, and deliver work. See the original link for full details.