OpenAI

OpenAI’s public incident record moved from identification at 09:04 UTC to mitigation at 09:06 and resolution at 09:28 on 31 July. The detailed update said some ChatGPT Business and Education users saw elevated errors when starting or continuing conversations. At this briefing’s fixed 09:23:13 window close, the incident was still in monitoring; the resolution arrived five minutes later and is included as an explicitly timed update. The visible status lifecycle lasted 24 minutes, but that is not necessarily the full customer-impact duration. OpenAI disclosed no failed component, cause, region, error rate, customer count or data-integrity effect.

OpenAI says it is contracting with Georgia Power for 3.2 gigawatts to be delivered in phases from 2028 to 2032 for Project Camellia in Effingham County. The scale is material, but it describes a future power-supply plan, not commissioned computing load today. Financing, design, phasing and operations remain unfinished.

OpenAI is a risk and accountability case because API and assistant-service availability now sits inside enterprise workflows, developer release paths, classrooms, support desks, public-service experiments, and regulated decision support. The public status record matters because customers need more than a reassurance that a service recovered; they need evidence that model-serving capacity, incident scope, notice quality, fallback design, and recovery proof can be audited like other cloud dependencies.

OpenAI’s compute procurement affects hyperscale cloud demand, data centre buildout, GPU supply chains and power infrastructure planning.

OpenAI is a leading AI platform company whose consumer adoption signals affect the AI application market.

OpenAI agrees to spend over $20bn on Cerebras chips to secure AI compute capacity and deepen infrastructure ties.

OpenAI acquires TBPN, marking a move into media to expand its role in shaping AI discussions and engaging technology audiences directly.

OpenAI announced a conditional Astral acquisition in March; by June, Charlie Marsh said the team had joined. The open-reference now matters more than an unproven rivalry narrative.

More than 100 organisations are calling for faster vulnerability repair, shared threat intelligence and wider access to defensive AI for critical infrastructure.

OpenAI’s eight-case field report shows where coding agents can shorten implementation work—and why scientific software teams still need hard acceptance criteria, realistic workloads and long-term owners.

OpenAI released o3-mini on 31 January 2025 as a smaller reasoning model for coding, mathematics and science. The release exposed low, medium and high reasoning-effort settings in the API and distributed the model across paid and free ChatGPT experiences with different controls. That made the useful product question more precise: not whether one model “reasons,” but which effort, tool, access tier and evaluation setup produced an acceptable answer.

OpenAI launched Deep Research in ChatGPT on 2 February 2025 as a Pro-only web capability for multi-step internet research. An early version of o3 could spend 5–30 minutes searching, reading and synthesizing online material and uploaded files into a cited report. That shifted the useful test from whether an answer sounded complete to whether a reviewer could trace its important claims to authoritative evidence, separate source material from inference and spot what the report missed.

OpenAI released Operator as a research preview on 23 January 2025 for ChatGPT Pro users in the United States. Its Computer-Using Agent could inspect screenshots and click, type or scroll in its own browser without a custom API for each website. That made the important output an external action rather than another answer—and made confirmations, takeover, supervision, error recovery and prompt-injection defenses part of the product's capability boundary.