Automated traffic mitigated with a human-verification check
Resolved - The human-verification layer is live and blocking automated traffic before it reaches the negotiation engine, and we registered the affected custom storefront domains with our verification provider so the check renders correctly for genuine shoppers. Negotiations, deals, discount codes, and vendor data were unaffected throughout.
9 Sep 2026, 16:20 UTC
Identified - We introduced a quick human-verification step to filter out the automated traffic. On storefronts served via a custom domain (rather than the default myshopify.com address), the interactive check needed the domain registered with our verification provider before it could display.
9 Sep 2026, 15:58 UTC
Investigating - We detected a pattern of automated (non-human) requests hitting the negotiation surface. Genuine negotiations continued running normally while we rolled out mitigation.
9 Sep 2026, 15:40 UTC
Backup model briefly served negotiations during an upgrade
Resolved - The primary model is fully restored and handling all live traffic, now with faster and more efficient responses from the upgrade. Automatic failover meant negotiations kept running throughout, and no negotiations, deals, discount codes, or vendor data were affected at any point.
3 Sep 2026, 19:12 UTC
Identified - We traced it to a response-format change in the primary model's upgrade. During the rollout a portion of negotiations were served by our backup model instead. Deals and discount codes continued to generate normally.
3 Sep 2026, 18:55 UTC
Investigating - While rolling out a performance and cost upgrade to the negotiation engine, we noticed some requests routing to our backup model. Automatic failover kept negotiations live while we investigated.
3 Sep 2026, 18:40 UTC
Elevated negotiation response times
Resolved - The upstream provider has confirmed recovery and traffic has returned to the primary model. No negotiations, deals, discount codes, or vendor data were affected at any point.
12 May 2026, 14:41 UTC
Monitoring - Automatic failover absorbed the spike and response times are back to normal. We are monitoring upstream recovery before routing traffic back.
12 May 2026, 14:22 UTC
Investigating - We detected a brief latency spike at an upstream AI provider. Affected requests automatically failed over to our backup model, so negotiations continued running.
12 May 2026, 14:20 UTC