OpenAI Multi-Service Outage (ChatGPT, Login, Access)

On 29 May 2026 OpenAI's public status history recorded four separate incidents the same day affecting ChatGPT conversations, login and account creation, ChatGPT access, and business subscription checkout. Teams routing production workloads through a single OpenAI dependency experienced correlated failure across inference, auth, and control-plane paths before OpenAI marked each incident fully recovered.

D8 · Vendor Resilience D4 · Observability D5 · Deployment Safety

What happened

OpenAI's status timeline for 29 May 2026 UTC logged conversation errors (~04:58), login and new-account creation errors (~04:57), a distinct business-plan checkout failure on web and mobile (~01:29), and later broad ChatGPT access issues (~13:28). Secondary user-impact spikes appeared on Downdetector and trade press the same day; OpenAI had not published a detailed root-cause postmortem on the public history page when PAI documented the episode. May 2026 already showed elevated warning-level events on prior days in independent uptime trackers cited by secondary outlets.

PSF Analysis

How the Production Safety Framework maps to this failure

A D8 failure pattern: multiple user-facing surfaces (conversations, login, checkout, access) degraded on the same day, exposing teams that treated a single OpenAI API key as the only production spine. D4 gaps appear when teams rely solely on status.openai.com while customer pain shows up earlier on internal error rates or social channels. D5 gaps appear when deployments assume inference availability equals full product availability without control-plane or auth fallbacks. Distinct from single-model latency spikes; belongs in vendor incident tracking for SLA reviews and failover rehearsal.

Controls that would have prevented this

Specific PSF controls mapped to each failure point

D8 · Vendor Resilience

Maintain a tested secondary model route (different vendor or region) with automated circuit breaking when OpenAI error budgets exceed policy.

D4 · Observability

Emit customer-visible SLO dashboards and synthetic probes on critical model routes; alert on your error budgets before public status moves to degraded.

D5 · Deployment Safety

Ship segmented degradation modes (read-only, cached responses, human handoff) when auth and inference paths fail together — do not assume API HTTP 200 means the product is up.

Outcome

OpenAI status history marked each listed 29 May 2026 incident as fully recovered (vendor wording). No detailed public root-cause postmortem on the history page at documentation time. Widely cited as a May 2026 vendor-resilience teachable moment for production AI teams.

vendor-outageopenaichatgptvendor-resilienceauth-failurecontrol-plane