OpenAI optimizes agentic workflows using WebSockets in the Responses API: reduced model latency and API overhead through connection-scoped caching. Improvements documented on the Codex agent loop.
OpenAI launches workspace agents in ChatGPT: Codex-powered agents that automate complex workflows, run in the cloud, and enable secure team collaboration across tools.
Hugging Face launches QIMMA, a quality-focused Arabic LLM leaderboard. The platform evaluates Arabic language models against rigorous criteria, providing a transparent benchmark for Arabic language performance.
OpenAI launches Codex Labs and partners with Accenture, PwC, and Infosys to help enterprises deploy and scale Codex across the software development lifecycle. The platform reaches 4M weekly active users.
OpenAI updates Codex app for macOS and Windows with computer use, in-app browsing, image generation, memory, and plugins. Expands use cases beyond code generation for developer workflows.
OpenAI releases GPT-Rosalind, a frontier reasoning model designed to accelerate drug discovery, genomics analysis, protein reasoning, and scientific research workflows.
OpenAI launches Trusted Access for Cyber program with leading security firms and enterprises using GPT-5.4-Cyber model. $10M in API grants allocated to strengthen global cybersecurity defense capabilities.
OpenAI updates Agents SDK with native sandbox execution and model-native harness, enabling developers to build secure, long-running agents with file and tool access.
OpenAI launches GPT-5.4-Cyber, a specialized cybersecurity model, to vetted defenders through its Trusted Access for Cyber program. The initiative strengthens safeguards as AI cybersecurity capabilities advance, balancing offensive and defensive applications.
OpenAI responded to an Axios supply chain attack by rotating macOS code signing certificates and updating applications. No user data was compromised, the company confirmed.
Hugging Face releases Waypoint-1.5, a 3D interactive simulation model optimized for consumer GPUs. The system generates high-fidelity controllable virtual worlds in real-time, reducing computational requirements compared to previous versions.
OpenAI launches pay-as-you-go pricing for Codex under ChatGPT Business and Enterprise plans, enabling teams to start and scale adoption with flexible billing.
Gradient Labs deploys AI agents powered by GPT-4.1 and GPT-5.4 mini/nano to automate banking support with low latency. Each customer gets a dedicated AI account manager. Production-ready approach for critical workflows.
Hugging Face trained language models on mRNA sequences across 25 species for $165. Models capture biological patterns and enable secondary structure and gene expression predictions without manual annotation.
Google DeepMind releases Gemini 3.1 Flash Live, an improved voice model with better precision and lower latency for more fluid, natural and precise voice interactions.
OpenAI releases Model Spec, a public framework defining expected behavior for its AI models. The document balances safety, user freedom, and accountability as AI systems advance.
OpenAI launches a Safety Bug Bounty program to identify AI abuse and security risks including agentic vulnerabilities, prompt injection, and data exfiltration. Researchers receive rewards for reporting discovered flaws.
OpenAI releases prompt-based teen safety policies for developers using gpt-oss-safeguard to moderate age-specific risks in AI systems.
OpenAI details its chain-of-thought monitoring approach for internal coding agents to detect misalignment. Analysis of real-world deployments aims to identify risks and strengthen AI safety safeguards.
OpenAI acquires Astral to accelerate Codex growth and power next-generation Python developer tools. The acquisition aims to integrate Astral's capabilities into OpenAI's developer tooling. No financial terms or timeline disclosed.
OpenAI outlines ChatGPT's defenses against prompt injection and social engineering in AI agent workflows. Approach: constraining risky actions and protecting sensitive data at the agent level.
Rakuten fixes code issues twice as fast using OpenAI's Codex model. The e-commerce company reports measurable productivity gains in developer bug-fixing workflows. Real-world deployment of generative AI for code assistance at enterprise scale.
Wayfair deploys OpenAI models to automate support ticket triage and enhance product attributes across millions of catalog items. Improves data accuracy and accelerates customer support response times at scale.
OpenAI introduces IH-Challenge, a training method that improves instruction hierarchy in frontier LLMs. It strengthens prioritization of trusted instructions, safety steerability, and resistance to prompt injection attacks.
Hugging Face introduces Storage Buckets on its Hub, enabling users to store and manage large files directly on the platform. This feature simplifies sharing of datasets, models, and artifacts without relying on external services.
OpenAI releases Codex Security in research preview: an AI application security agent that analyzes project context to detect, validate, and patch complex vulnerabilities with higher confidence and reduced false positives.
Descript leverages OpenAI's reasoning models to automate multilingual video dubbing at scale while preserving timing and meaning of original content.
Hugging Face releases a comprehensive guide for deploying robotics AI models on embedded platforms. The project covers dataset recording, Vision Language Action (VLA) fine-tuning, and on-device optimizations to reduce latency and memory consumption.
OpenAI introduces CoT-Control, a tool measuring reasoning models' ability to control their chains of thought. Results show these models struggle to direct their internal processes, reinforcing monitorability as a critical AI safety safeguard.
Hugging Face introduces Modular Diffusers, a modular architecture for diffusion pipelines. Reusable components enable combining schedulers, denoisers, and other building blocks to create custom image generation models without rewriting code.
OpenAI launches ChatGPT for Excel with new financial data integrations powered by GPT-5.4 to accelerate modeling, research, and analysis in regulated environments.
OpenAI releases GPT-5.3 Instant, a model optimized for everyday conversations with reduced latency and improved fluidity. Positioned as a faster, lighter alternative to previous versions for routine use.
OpenAI announces a contract with the U.S. Department of Defense outlining safety red lines, legal protections, and deployment protocols for AI systems in classified environments. Specific terms remain confidential but establish ethical boundaries for military applications.
OpenAI and Amazon announce strategic partnership: OpenAI's Frontier platform integrates with AWS, expanding AI infrastructure, custom models, and enterprise AI agents.
Amazon Bedrock introduces a stateful runtime environment for agents, providing persistent orchestration, memory, and secure execution for multi-step AI workflows powered by OpenAI.
OpenAI and Pacific Northwest National Laboratory introduce DraftNEPABench, a benchmark measuring AI coding agents' ability to accelerate federal permitting. Tests show potential 15% reduction in NEPA drafting time and modernized infrastructure reviews.
A new paper investigates AI agent reliability by quantifying the gap between claimed capabilities and actual performance. The study proposes methods to measure this divergence and improve the robustness of agent systems.
OpenAI commits $7.5M to The Alignment Project for independent AI alignment research. Aims to strengthen global efforts on AGI safety and security risks.
IBM and UC Berkeley introduce IT-Bench and MAST to diagnose enterprise agent failures. IT-Bench benchmarks agents on realistic IT tasks, while MAST (Multi-Agent Simulation Testbed) simulates complex environments to test multi-agent system robustness.
Recraft V4 generates art-directed images with strong composition, accurate text rendering, and design taste. The model produces editable SVGs. Four variants now available on Replicate.