Ce que OpenAI a publié en avril 2026

Introducing Advanced Account Security

OpenAI introduced Advanced Account Security, offering phishing-resistant login, stronger recovery options, and enhanced protections to prevent account takeover and safeguard sensitive data. The featur…

GPT‑5.5 Instant: smarter, clearer, and more personalized

OpenAI replaced ChatGPT’s default model with GPT-5.5 Instant, delivering 52.5% fewer hallucinations in high-stakes domains like medicine and law, 37.3% fewer factual errors on challenging prompts, and…

Scaling Trusted Access for Cyber with GPT‑5.5 and GPT‑5.5‑Cyber

OpenAI introduced GPT-5.5-Cyber in limited preview, offering specialized cybersecurity workflows with stronger verification and account controls. GPT-5.5 with Trusted Access for Cyber (TAC) remains th…

OpenAI Developers plugin for Codex

OpenAI releases a new plugin for Codex aimed at developers, enhancing integration capabilities. The plugin is designed to streamline workflows for developers using Codex.

Where the goblins came from

OpenAI detailed the emergence of 'goblin outputs'—personality-driven quirks in GPT-5 behavior—tracing their spread via training data and model interactions. The company identified root causes, includi…

GPT‑5.5 Instant System Card

OpenAI introduced GPT-5.5 Instant, the first Instant model classified as High capability in Cybersecurity and Biological & Chemical Preparedness categories, requiring stricter safeguards. The model is…

Building the compute infrastructure for the Intelligence Age

OpenAI announced the expansion of its Stargate compute infrastructure to meet rising AI demand, adding new data center capacity aimed at powering AGI. The move underscores the company's commitment to …

OpenAI models, Codex, and Managed Agents come to AWS

OpenAI announced the availability of its GPT models, Codex, and Managed Agents on AWS, allowing enterprises to deploy secure AI solutions within their AWS environments. This integration expands OpenAI…

OpenAI introduced Advanced Account Security, an opt-in feature requiring passkeys or physical security keys for sign-in and disabling password-based login. It tightens account recovery, shortens sessi

Business ChatGPT & Codex

Business ChatGPT & Codex A secure, collaborative workspace for startups and growing businesses / user / month Get started What's included: Everything in ChatGPT Plus and Business Codex plans Unli…

OpenAI available at FedRAMP Moderate

OpenAI announced FedRAMP Moderate authorization for ChatGPT Enterprise and the OpenAI API, allowing secure AI adoption by U.S. federal agencies. This compliance removes a major barrier for government …

An open-source spec for Codex orchestration: Symphony

OpenAI released Symphony, an open-source agent orchestrator that turns issue trackers like Linear into control planes for coding agents. The system increased landed PRs by 500% in some teams by decoup…

The next phase of the Microsoft OpenAI partnership

OpenAI and Microsoft announced an amended partnership agreement to simplify their collaboration and provide long-term clarity for continued AI innovation at scale. The changes aim to strengthen their …

An open-source spec for orchestration: Symphony

OpenAI introduced Symphony, an open-source specification designed to transform issue trackers into always-on agent systems. This aims to enhance engineering productivity by reducing context switching …

Choco automates food distribution with AI agents

Choco, a B2B food distribution platform, deployed OpenAI’s APIs to automate manual processes, reduce operational friction, and improve productivity in its supply chain. The integration enabled faster …

How OpenAI delivers low-latency voice AI at scale

OpenAI redesigned its WebRTC infrastructure to reduce latency in real-time voice AI by splitting media routing (via a lightweight relay) from protocol termination (handled by a stateful transceiver). …

OpenAI and Microsoft announced an amended partnership agreement to simplify their collaboration, providing long-term clarity and flexibility. The deal enables joint scaling of AI infrastructure, inclu

Introducing GPT-5.5

OpenAI introduced GPT-5.5, claiming it is their most advanced model to date, optimized for complex tasks such as coding, research, and data analysis. The model is designed to be faster and more capabl…

Our commitment to community safety

OpenAI announced expanded safeguards for ChatGPT to detect and prevent real-world violence, including refined model training, automated detection systems, and human review processes. The company now e…

OpenAI and AWS expanded their partnership to bring OpenAI’s frontier models (including GPT-5.5), Codex, and Managed Agents to Amazon Bedrock. Enterprises can now deploy these AI capabilities within th

Making ChatGPT better for clinicians

OpenAI introduced ChatGPT for Clinicians, a free version of ChatGPT tailored for U.S. physicians, nurse practitioners, physician assistants, and pharmacists to assist with documentation, medical resea…

GPT‑5.5 Bio Bug Bounty

OpenAI introduced a Bio Bug Bounty program for GPT-5.5, inviting researchers to find universal jailbreaks that bypass its five-question bio safety challenge. Applications are open until June 22, 2026,…

GPT-5.5 System Card

OpenAI introduced GPT-5.5, a model optimized for complex real-world tasks like coding, research, and document creation. It demonstrates improved task understanding, reduced need for user guidance, bet…

GPT‑5.5 System Card

OpenAI launched GPT-5.5, a model designed for complex real-world tasks like coding, research, and document creation, with improved task understanding and tool integration. It includes stronger safegua…

GPT-5.5 Bio Bug Bounty

OpenAI introduced the GPT-5.5 Bio Bug Bounty, a red-teaming initiative offering up to $25,000 for discovering universal jailbreaks that could bypass bio safety safeguards in GPT-5.5. The program aims …

Introducing GPT‑5.5

OpenAI released GPT-5.5, a new AI model surpassing GPT-5.4 in intelligence, efficiency, and agentic workflows. It excels in coding, knowledge work, and scientific research, achieving state-of-the-art …

Pricing change detected for Openai

Pricing updated for Openai: - Free: now includes Limited access to flagship model GPT-5.3, Limited messages and uploads, Limited and slower image generation, Limited deep research, Limited memory and …

What Parameter Golf taught us

OpenAI’s Parameter Golf challenge drew over 1,000 participants and 2,000 submissions, emphasizing the role of AI coding agents in lowering barriers to ML research. The event revealed talent, fostered …

ChatGPT Images 2.0 System Card

OpenAI introduced ChatGPT Images 2.0, a major upgrade to its image generation model featuring enhanced world knowledge, instruction following, and dense text generation. The model now includes a 'thin…

OpenAI helps Hyatt advance AI among colleagues

Hyatt has deployed ChatGPT Enterprise across its global workforce, leveraging GPT-5.4 and Codex to enhance productivity, streamline operations, and improve guest experiences. The integration aims to a…

Codex for (almost) everything

OpenAI released an updated Codex app for macOS and Windows, introducing computer use, in-app browsing, image generation, memory, and plugins to enhance developer workflows. These additions aim to stre…

Introducing workspace agents in ChatGPT

OpenAI introduced workspace agents in ChatGPT, enabling teams to create shared AI agents for complex, long-running workflows. Powered by Codex, these agents operate within organizational permissions, …

GPT‑5.3 Instant: Smoother, more useful everyday conversations

OpenAI released GPT-5.3 Instant, an updated version of its most-used model, focusing on improving everyday conversational quality by reducing unnecessary refusals, refining tone, and enhancing context…

Introducing GPT-Rosalind for life sciences research

OpenAI launched GPT-Rosalind, a frontier reasoning model designed to accelerate drug discovery, genomics analysis, protein reasoning, and scientific research workflows. The model leverages advanced re…

Introducing OpenAI Privacy Filter

OpenAI launched Privacy Filter, an open-weight model designed to detect and redact personally identifiable information (PII) in text with high accuracy. The model runs locally, supports context-aware …

Accelerating the cyber defense ecosystem that protects us all

OpenAI announced the launch of its Trusted Access for Cyber initiative, partnering with leading security firms and enterprises to deploy GPT-5.4-Cyber for cyber defense. The program includes $10 milli…

OpenAI released a major update to Codex, expanding its capabilities beyond coding to include computer control, image generation, and workflow automation. Codex can now operate a user's computer, inter

The next evolution of the Agents SDK

OpenAI introduced a major update to its Agents SDK, adding native sandbox execution and a model-native harness. These changes enable developers to build secure, long-running agents that can operate ac…

Pricing updated for Openai: - Go: new promotion — This plan may include ads - Go: now includes Expanded access - Plus: allowance changed from Unlimited subject to abuse guardrails to Advanced access -

OpenAI launched the Trusted Access for Cyber initiative, committing $10 million in API credits to support cybersecurity defenders through its Cybersecurity Grant Program. The program provides advanced

Trusted access for the next era of cyber defense

OpenAI expanded its Trusted Access for Cyber program by introducing GPT-5.4-Cyber, a new AI model tailored for vetted cybersecurity defenders. The update strengthens safeguards as AI-driven cybersecur…

Enterprises power agentic workflows in Cloudflare Agent Cloud with OpenAI

Cloudflare has integrated OpenAI’s GPT-5.4 and Codex models into its Agent Cloud platform, enabling enterprises to build, deploy, and scale AI agents for real-world tasks with enhanced speed and secur…

Analyzing data with ChatGPT

OpenAI has introduced a new educational initiative called 'ChatGPT Academy' focused on data analysis, enabling users to explore datasets, generate insights, create visualizations, and convert findings…

Creating images with ChatGPT

OpenAI has introduced a new educational resource focused on image generation using ChatGPT. The content emphasizes teaching users how to create and refine images through clear prompts, iterative desig…

ChatGPT for marketing teams

OpenAI highlights how marketing teams leverage ChatGPT to streamline campaign planning, content generation, performance analysis, and execution. The platform enables faster ideation and implementation…

AI fundamentals

The webpage provides a beginner-friendly explanation of artificial intelligence (AI), detailing how it functions and its applications through tools like ChatGPT, which utilize large language models. T…

ChatGPT for customer success teams

OpenAI’s Academy page highlights how customer success teams leverage ChatGPT to enhance account management, streamline communication, reduce customer churn, and drive product adoption and renewals. Th…

Prompting fundamentals

The webpage introduces a learning resource focused on prompting fundamentals for ChatGPT, emphasizing the importance of writing clear and effective prompts to enhance response quality. It targets user…

Using projects in ChatGPT

OpenAI has introduced a new feature called 'Projects' within ChatGPT, designed to help users organize chats, files, and instructions more efficiently. This tool enables users to manage ongoing work an…

Writing with ChatGPT

OpenAI introduces a new resource under its Academy program focused on leveraging ChatGPT for writing tasks. The content emphasizes using ChatGPT to assist in drafting, revising, and refining written c…

OpenAI is scaling its Trusted Access for Cyber (TAC) program to thousands of verified cyber defenders and hundreds of teams, introducing a new cyber-permissive model variant, GPT-5.4-Cyber, fine-tuned

Using skills

OpenAI introduces a new framework for creating and deploying 'ChatGPT skills,' which are reusable workflows designed to automate repetitive tasks and standardize outputs. These skills enable users to …

Scaling Codex to enterprises worldwide

OpenAI launched Codex Labs and partnered with seven global systems integrators (GSIs) including Accenture, Capgemini, and TCS to accelerate enterprise adoption of Codex. Over 4 million developers now …

Pricing updated for Openai: - Plus: added annual billing at $16/mo (20% off) - Plus: now includes Unlimited subject to abuse guardrails - Pro: $200/mo → $25/mo (88% decrease) - Pro: added annual billi

Announcing the OpenAI Safety Fellowship

OpenAI has launched a pilot program called the Safety Fellowship to support independent research in AI safety and alignment, aiming to cultivate the next generation of talent in these critical areas. …

Introducing the Child Safety Blueprint

OpenAI introduced a Child Safety Blueprint to combat AI-enabled child sexual exploitation, outlining a framework for strengthening U.S. child protection standards in the age of AI. The blueprint prior…

Industrial policy for the Intelligence Age

OpenAI introduces a policy framework advocating for an 'industrial policy' centered on human-centric principles to guide the development and deployment of advanced intelligence systems. The proposal e…

The next phase of enterprise AI

OpenAI has entered a new phase of enterprise AI adoption, with enterprise revenue now exceeding 40% of total revenue and on track to match consumer revenue by the end of 2026. The company highlights r…

OpenAI acquires TBPN

OpenAI has acquired TBPN, a move aimed at accelerating global conversations around artificial intelligence while supporting independent media. The acquisition is designed to expand dialogue with AI bu…

Codex now offers more flexible pricing for teams

OpenAI has introduced pay-as-you-go pricing for its Codex product within ChatGPT Business and Enterprise plans, offering teams a more flexible alternative to traditional subscription models. This chan…

OpenAI released a major update to its Agents SDK, introducing a model-native harness and native sandbox execution to simplify agent development for long-horizon tasks. The new capabilities allow agent

Gradient Labs gives every bank customer an AI account manager

Gradient Labs has integrated OpenAI’s GPT-4.1 and GPT-5.4 mini and nano models to power AI agents designed for automating banking support workflows. The focus is on delivering low-latency and high-rel…