AI Systems Status Report - February 2026

Summary

This report covers the operational status of major AI systems during February 2026, including Anthropic, Cohere, DeepSeek, Google Gemini, Groq Cloud, OpenAI, Perplexity, Replicate, and xAI. The data includes official incidents reported on vendor status pages and unconfirmed incidents detected through IsDown's monitoring systems.

Important: Each provider has their own framework for reporting incidents. A provider showing more incidents does not mean they are less reliable - it may indicate more transparent reporting. These providers also vary significantly in size and market share, making direct comparisons inappropriate.

Provider	Official Incidents	Total Incident Time	Peak Incident Day	Typical Resolution	Unconfirmed Incidents
Anthropic	55 (16 major, 39 minor)	~282.2 hours	Tuesday	30-60 min	0
Cohere	1 (0 major, 1 minor)	~1.3 hours	Friday	1-2 hours	0
DeepSeek	3 (0 major, 3 minor)	~3.1 hours	Friday	< 15 min / 30-60 min / 2-4 hours	0
Google Gemini	1 (0 major, 1 minor)	~86.4 hours	Thursday	4+ hours	1
Groq Cloud	2 (0 major, 2 minor)	~3.9 hours	Thursday and Saturday	1-2 hours / 2-4 hours	0
OpenAI	23 (1 major, 22 minor)	~80.2 hours	Wednesday	2-4 hours	0
Perplexity	2 (1 major, 1 minor)	~1.2 hours	Thursday and Monday	15-30 min / 30-60 min	0
Replicate	0	-	-	-	0
xAI	0	-	-	-	0

Anthropic

Anthropic reported 55 incidents (16 major, 39 minor) totaling approximately 282.2 hours of incident time. The average resolution time was 307 minutes per incident.

The majority of incidents involved elevated error rates across various Claude models, particularly Opus 4.5, Opus 4.6, Sonnet 4.5, and Sonnet 4.6. Multiple incidents affected API connectivity and model availability throughout the month. Several platform-specific issues were reported, including problems with Claude Desktop app functionality, SSO authentication, and user-created skills on claude.ai. Credit purchase and usage reporting systems also experienced disruptions. A notable long-duration incident involved intermittent errors in skills-related functionality that persisted for over 159 hours.

Cohere

Cohere reported 1 incident (0 major, 1 minor) totaling approximately 1.3 hours of incident time. The average resolution time was 77 minutes.

The single incident affected chat models, causing unresponsiveness and timeouts for approximately 78 minutes on February 13th.

DeepSeek

DeepSeek reported 3 incidents (0 major, 3 minor) totaling approximately 3.1 hours of incident time. The average resolution time was 61 minutes per incident.

All three incidents involved degraded performance affecting both the DeepSeek web interface and API services. Two incidents occurred on February 6th, with one lasting approximately 145 minutes, while another incident on February 3rd lasted about 31 minutes.

Google Gemini

Google Gemini reported 1 incident (0 major, 1 minor) totaling approximately 86.4 hours of incident time. The resolution time was 5,181 minutes.

The single incident involved conversation history visibility issues affecting Gemini Web and mobile apps. The incident began on February 18th and lasted over 86 hours while engineering teams worked to restore missing user conversation metadata caused by a background process issue.

Unconfirmed Incidents

Note: Unconfirmed incidents are potential issues detected by IsDown but never officially acknowledged by the vendor. These serve as indicators only and do not confirm an actual outage occurred.

IsDown detected 1 unconfirmed incident on February 13th based on user reports indicating a possible outage.

Groq Cloud

Groq Cloud reported 2 incidents (0 major, 2 minor) totaling approximately 3.9 hours of incident time. The average resolution time was 115 minutes per incident.

Both incidents involved degraded performance specifically affecting the meta-llama/llama-4-scout-17b-16e-instruct model, with users experiencing elevated error rates and slow responses. The incidents occurred on February 5th and 7th, lasting approximately 120 and 112 minutes respectively.

OpenAI

OpenAI reported 23 incidents (1 major, 22 minor) totaling approximately 80.2 hours of incident time. The average resolution time was 209 minutes per incident.

Incidents affected various OpenAI services throughout the month. Multiple incidents involved elevated error rates for ChatGPT and platform services, with several affecting conversation functionality and login processes. Specific model issues included problems with GPT 5.2, text-embedding-3-small, and image generation services. Other notable incidents affected Codex Cloud (including GitHub dependency issues), Custom GPT updates, ChatGPT Go, Sora 2, support email processing, shopping research features, and artifact generation. One particularly long-duration incident involved image generation errors lasting over 20 hours.

Perplexity

Perplexity reported 2 incidents (1 major, 1 minor) totaling approximately 1.2 hours of incident time. The average resolution time was 35 minutes per incident.

The incidents included website stability issues with elevated error rates lasting approximately 20 minutes on February 5th, and a Sonar API incident lasting about 52 minutes on February 16th.

Replicate

Replicate reported no incidents during February 2026.

xAI

xAI reported no incidents during February 2026.

Conclusion

February 2026 saw varying levels of incident activity across AI service providers. Anthropic reported the highest number of incidents at 55, while Replicate and xAI reported no incidents. Total incident durations ranged from 1.2 hours to 282.2 hours across providers that experienced issues. Resolution times varied significantly, from an average of 35 minutes for Perplexity to 307 minutes for Anthropic. The most common incident types across all providers involved elevated error rates, model-specific performance degradation, and platform functionality issues.