ChatGPT and Codex Hit by Major Outage Affecting Thousands of Users Worldwide

By Hector Herrera | April 20, 2026

OpenAI's ChatGPT and Codex went down for a significant stretch of Monday morning, knocking out 12 service components and disrupting tens of thousands of users across the UK and United States. The incident is a reminder that centralized AI infrastructure carries real operational risk — and that most businesses still have no fallback plan when it fails.

What Happened

The outage began at 10:05am ET on April 20, according to TechRadar's live coverage. Twelve service components failed simultaneously, including:

• Core conversations — the primary ChatGPT chat interface
• User login — locking users out of their accounts entirely
• Voice mode — the audio-in, audio-out feature used for hands-free and accessibility workflows
• Image generation — DALL·E-powered image creation
• Codex — OpenAI's AI-assisted coding platform embedded in developer pipelines

Disruption tracking platform Downdetector recorded peak reports of over 8,700 incidents in the UK and approximately 1,900 in the United States. OpenAI confirmed the incident publicly. The company deployed a fix and reported it was monitoring recovery by 12:48pm ET — roughly two hours and forty minutes after the outage began.

Why It Matters

ChatGPT has moved well beyond novelty. Enterprises have baked it into knowledge work, customer support, and content pipelines. Codex specifically sits inside the active coding workflows of development teams who rely on AI pair-programming — not as a convenience, but as a daily tool. When it goes dark mid-session, work stops.

The UK absorption numbers are notable. More than 8,700 disruption reports is a substantial volume for a non-consumer product primarily used during business hours. It signals that European enterprise adoption of ChatGPT has grown to a point where a US-originating outage registers at scale across the Atlantic.

Login failures compound the impact. When users can't authenticate at all — not just use features — it effectively treats the platform as fully down for affected accounts, even if some backend systems remain operational.

The Deeper Issue: AI Vendor Concentration

This outage won't be the last. OpenAI runs one of the most heavily trafficked AI platforms on the planet, and no infrastructure at this scale avoids downtime. The real risk isn't a single three-hour window — it's that most organizations have built AI-dependent workflows with no redundancy layer.

Running a secondary model provider — whether that's Anthropic's Claude, Google's Gemini, or another — isn't over-engineering. It's the same logic that pushed IT teams toward multi-cloud strategies after early AWS outages. The AI equivalent of that conversation is overdue.

What to Watch

OpenAI has not published a post-incident root cause analysis as of this writing. Monitor the OpenAI status page for a formal incident report. If the cause was infrastructure-level rather than a code deployment, expect renewed scrutiny of OpenAI's reliability commitments in enterprise contract negotiations.

More broadly: watch whether this incident accelerates enterprise demand for AI service-level agreements (SLAs) with financial teeth — something most current AI vendor contracts conspicuously lack.

Source: TechRadar — ChatGPT down, April 2026