engblogs

Technical overview of the rollout for ChatGPT ads, detailing CPC bidding, a beta self-serve Ads Manager, broader partner support, and privacy-preserving measurement to help advertisers buy, manage, and optimize campaigns.

GPT-5.5 Instant introduces smarter, more personalized responses with improved factuality and context reuse across chats, files, and Gmail, powered by memory sources and clearer, more concise answers.

A technical overview of Europe's AI-driven youth safety framework and the EMEA grants, outlining age-appropriate safeguards, age assurance, and practical governance for learning, wellbeing, and responsible AI use in education.

Explores GPT-5.5 Instant as a high-capability Instant model, detailing its safety mitigations, safeguards, and role in Cybersecurity and Biological & Chemical Preparedness, with baselines set against GPT-5.3 Instant.

Unveiling MRC (Multipath Reliable Connection), a multi-plane, SRv6-based, packet-spraying network protocol that delivers resilient, scalable AI training across thousands of GPUs.

Overview of the GPT-5.5 Instant rollout, highlighting enhanced personalization, improved factuality, memory sources with transparent context, and expanded availability across ChatGPT plans and API.

A technical overview of GPT-5.5 Instant, detailing its high-capability safety safeguards, Cybersecurity and Biological & Chemical Preparedness framing, and its naming relations to GPT-5.5 Thinking and GPT-5.3 Instant.

OpenAI delivers low-latency voice AI at scale by a split relay and transceiver WebRTC architecture that minimizes the public UDP surface, preserves standard WebRTC semantics, and routes the first packet to the owning transceiver for near-instant, 1:1 conversations.

PwC and OpenAI are turning the CFO into an agentic, governance-driven operating model by deploying AI agents that automate finance workflows, coordinate across systems, surface risks, and enable proactive, governed decisions.

A scalable WebRTC-based architecture for OpenAI's voice AI that separates signaling/termination from media routing via a lightweight relay and a stateful transceiver, reducing the public UDP footprint while delivering low-latency, 1:1 real-time audio.

OpenAI and PwC pair AI agents with core finance workflows to automate planning, forecasting, and close, connect systems across procurement, payments, treasury, and tax, and embed governance and human oversight to scale agentic finance for the CFO.

OpenAI delivers low-latency voice AI at scale by separating WebRTC session termination from routing: a transceiver handles state while a lightweight global relay forwards packets, shrinking the public UDP surface and dramatically reducing first-packet latency.