Latest publications
OpenAI model refuted the central hypothesis of discrete geometry
An OpenAI neural network solved the unit distance problem, an 80-year-old problem in discrete geometry, refuting a central hypothesis and opening new possibilities for AI in mathematics.
OpenAI and Dell: Codex now runs on enterprises' own servers
OpenAI and Dell are joining forces so enterprises can deploy the AI coder Codex on their own servers without sending data to the cloud — a new standard for critical environments.
OpenAI and Malta provide all citizens with ChatGPT Plus access and AI training
OpenAI and Malta launched a program that gives all citizens free access to ChatGPT Plus, including training in practical AI skills and responsible use.
OpenAI explained how Codex helps sales teams prepare briefs, forecasts, and plans
OpenAI released a practical guide to Codex for sales teams: the tool gathers data from CRM, emails, and calls and turns it into briefs, forecasts, and account plans.
OpenAI showed how Codex helps business operations teams prepare materials for leadership decisions
OpenAI showed how Codex helps business operations teams turn data from trackers, KPI data, and notes into ready briefs, leadership updates, and scenario models.
OpenAI explained how it rebuilt WebRTC for low-latency voice AI
OpenAI showed how a new relay + transceiver stack for ChatGPT Voice and the Realtime API cuts latency and removes the need to keep thousands of public UDP ports open.
OpenAI and PwC join forces to automate finance and reshape the CFO role
OpenAI and PwC have announced a partnership aimed at helping companies introduce AI agents into finance: automate processes, improve forecasting accuracy, and strengthen internal controls.
OpenAI launches Ads Manager for ChatGPT and a pay-per-click model for ads
OpenAI is expanding advertising in ChatGPT: the company has opened a beta version of a self-serve Ads Manager, added pay-per-click pricing and new metrics, while preserving privacy and keeping ads separate from conversat
OpenAI introduced MRC — a network protocol for 100,000-GPU AI training clusters
Through the Open Compute Project, OpenAI released the MRC protocol, which distributes traffic across hundreds of routes and helps large AI clusters ride out failures without a noticeable training slowdown.
OpenAI: leading companies make 3.5x deeper use of AI and deploy Codex more actively
In its B2B Signals report, OpenAI showed that leading companies use AI not just more often, but more deeply: they have 3.5x more “intelligence” per employee, while Codex usage is 16x higher.
Uber and OpenAI roll out AI assistants and voice ride booking for drivers and passengers
Uber uses OpenAI models to tell drivers where and when they can earn more, while letting passengers book rides faster by voice directly in the app.
Singular Bank implemented ChatGPT and Codex: bankers save up to 90 minutes a day
Singular Bank launched its internal assistant Singularity, based on ChatGPT and Codex: it speeds up portfolio analysis, meeting preparation, and follow-up, saving bankers up to 90 minutes a day.
Parloa launched voice AI agents to support large companies using OpenAI models
Parloa showed how it builds voice AI agents for customer support using OpenAI models: the AMP platform lets teams model dialogues, check quality, and run the service live.
OpenAI launches Trusted Contact in ChatGPT for suicide-risk notifications
OpenAI has begun rolling out Trusted Contact in ChatGPT, an optional feature that can notify a preselected loved one when there are signs of a serious suicide risk.
OpenAI adds GPT-Realtime-2, Translate, and Whisper to the API for voice applications
OpenAI released three voice models for the API: one can reason in real time, another translates speech on the fly, and the third provides streaming transcription for calls and meetings.
OpenAI expanded Trusted Access for Cyber and added GPT-5.5-Cyber for defenders
OpenAI expanded Trusted Access for Cyber, opening GPT-5.5 and GPT-5.5-Cyber to verified defenders for faster vulnerability discovery and protection of critical infrastructure.
OpenAI: Simplex cut screen development time by 70% after adopting Codex
OpenAI published a Simplex case study: the company shifted development to ChatGPT Enterprise and Codex and recorded faster design, development, and integration testing.
OpenAI unveils teen protection plan in Europe and €500,000 in EMEA grants
OpenAI introduced the European Youth Safety Blueprint and selected 12 recipients of €500,000 in EMEA grants for AI literacy, mental health, and teen protection projects.
OpenAI explained what data ChatGPT uses for training and how it protects privacy
OpenAI released a clear explanation of how ChatGPT is trained on data, how personal information is filtered, and which settings let users disable the use of new chats for training.
OpenAI details Codex safeguards: sandboxes, approvals, and audits of agent actions
OpenAI described the restrictions and logs it uses for Codex: environment isolation, auto-approval of safe actions, network rules, and telemetry for security teams.
OpenAI: how large companies scale AI through trust, control, and quality
OpenAI analyzed the experience of Philips, BBVA, Mirakl, Scout24, JetBrains, and Scania and showed that scaling AI depends not on the model, but on trust, workflow, and quality control.
OpenAI launched DeployCo — a company for deploying AI into business workflows
OpenAI spun AI deployment into a separate unit, DeployCo, assigning about 150 engineers and $4 billion to it to move AI projects from pilots into real business systems faster.
OpenAI: ChatGPT audience grew sharply among users aged 35+ in early 2026
In the first quarter of 2026, ChatGPT made sharp gains in the mass market: usage grew fastest among people over 35, and the gender balance became noticeably more even.
OpenAI sums up Parameter Golf: how coding agents are changing machine learning research
OpenAI shared what it learned from the Parameter Golf competition: more than 1,000 participants and 2,000 submissions showed that AI agents speed up ML experiments, but make verification and evaluation of results more di
AutoScout24 accelerated development 10x with OpenAI's Codex and ChatGPT
AutoScout24 rolled out ChatGPT to 2,000 employees and Codex to 1,000 product and engineering roles, cutting some project timelines from 2–3 weeks to 2–3 days.
OpenAI showed how finance teams use Codex for reporting and scenario planning
OpenAI released a practical guide on how finance teams use Codex for MBRs, model checks, report packs, variance bridge, and scenario planning.
OpenAI explained the TanStack npm attack and asked users to update macOS apps by June 12
After the attack on the TanStack library, OpenAI found no signs of a user data leak, but is rotating signing certificates and requires macOS apps to be updated by June 12, 2026.
OpenAI showed how it built a secure Codex sandbox for Windows
OpenAI revealed the architecture of Codex's Windows sandbox: separate sandbox users, firewall rules, and limited write access let the agent work locally without Full Access.
OpenAI updated ChatGPT to more accurately detect risk in sensitive conversations
OpenAI outlined new safeguards for ChatGPT: the model now does a better job of taking previous messages and separate chats into account to spot signs of risk in time and respond more cautiously.
Databricks integrated GPT-5.5 into enterprise AI agents after a record score on OfficeQA Pro
Databricks made GPT-5.5 available for customer agentic scenarios: the model was the first to exceed 50% accuracy on OfficeQA Pro and cut errors by 46% versus GPT-5.4.