GitHub Down: Copilot, Actions and API Hit by Worldwide Outage

Published: 17 Aug 2026, 10:33 pm IST

The GitHub outage began on the morning of August 17, 2026, and as of 21:12 IST (15:42 UTC) it is still unresolved. Five services — API Requests, Issues, Pull Requests, Actions and Copilot — are marked as a major outage on GitHub’s own status page. Git Operations, Webhooks and Pages are degraded.

If you are seeing 500 errors, failed pushes, stalled CI runs or a dead Copilot, the problem is not on your machine.

What GitHub has confirmed is down

GitHub acknowledged the incident at 9:40 AM EDT, initially reporting performance problems on some services. Within an hour the failure had spread across most of the platform.

The company’s status page puts the damage in numbers:

Affected area	Error rate
Web experiences and API traffic	~20%
Archive downloads and raw repository content	~50%
SAML and OIDC authentication, SCIM, Team Sync	Impacted

A 20% error rate means roughly one in every five requests to GitHub.com or the API is failing. For raw content and archive downloads, close to half of all requests are failing — which is why package installs and Docker builds that pull from GitHub have been breaking as much as the website itself.

At 11:42 AM EDT, GitHub said it had begun applying mitigations. Error rates have not moved since.

Copilot down, but the AI models are fine

This detail matters and most coverage has skipped it.

Copilot is listed under major outage, but Copilot AI model providers remain operational on GitHub’s status page. The models answering your prompts are up. The failure sits in GitHub’s own service layer between your editor and those models.

That is why the query “github copilot down” registered as a breakout search term today alongside “copilot down,” which spiked over 350%. Developers noticed Copilot dying before they noticed the website was struggling, because most people have Copilot open in an IDE all day but only visit github.com a few times.

Scale of the disruption

Downdetector logged more than 3,000 reports, with regional peaks of roughly 2,100 in the United States and 849 in India. Reports started climbing around 6:40 PM IST — right in the middle of the evening working window for Indian development teams and the start of the US business day.

That overlap is why this one hurt more than a typical GitHub wobble. Indian IT services teams, startup engineering teams and US-based product teams were all mid-shift when repositories stopped loading.

What is actually broken for you

Based on the affected components, here is what will and will not work:

Likely failing

Loading repositories, commits and pull request pages
Opening or commenting on issues
GitHub Actions workflow runs and CI/CD pipelines
API calls from any tool that integrates with GitHub
git archive downloads, raw file fetches, release asset downloads
Enterprise SSO logins via SAML or OIDC
Copilot completions and chat
Webhook deliveries to your services

Likely still working

Local git commands — commit, branch, merge, rebase, stash
Packages and Codespaces, listed as operational
Any existing local clone of your repository

The single most useful thing to remember: git is distributed. Your full repository history lives on your laptop. Nothing you have already committed is at risk because GitHub is down.

What developers should do right now

Keep committing locally. Commit as normal and push when service returns. There is no reason to stop working.

Do not retry pushes in a loop. Repeated failed pushes add load to a system already applying mitigations, and a partially failed push during an unstable window is a worse outcome than waiting.

Pause non-urgent CI/CD. If your pipeline pulls dependencies from GitHub releases or raw content, a 50% error rate means roughly half your builds will fail for reasons unrelated to your code. Failed builds during an outage pollute your build history and can trigger false alerts.

Check your dependency chain. GitHub is upstream of a huge amount of infrastructure. If your deployment pulls Go modules, Homebrew formulae, GitHub Actions marketplace actions or install scripts hosted on GitHub, your unrelated service can fail because of this outage.

Do not push a hotfix during an outage. If you are mid-incident on your own product, a partially propagated deploy is harder to debug than a delayed one.

Watch the primary source. githubstatus.com is the authoritative feed. Third-party monitors were reporting conflicting states during this incident — some showed Copilot as operational while GitHub’s own page had it in major outage.

What caused the GitHub Down outage?

GitHub has not published a cause. Its investigation is ongoing and no root cause analysis has been released.

Speculation is circulating that this is an Azure problem, since GitHub runs on Microsoft infrastructure. GitHub has not attributed the incident to any upstream provider, and no Azure link has been confirmed. Treat that claim as unverified until GitHub publishes its RCA.

One data point does argue against a simple upstream failure: Copilot’s model providers stayed operational while Copilot itself went down. That pattern points inward, toward GitHub’s own service layer, rather than to a shared dependency taking everything down at once.

Is this happening more often?

GitHub’s 90-day uptime figures on its own status page tell a mixed story. Actions sits around 99.36%, API Requests near 99.84%, and Issues and Pull Requests around 99.91%.

For context, 99.36% uptime over 90 days is roughly 14 hours of downtime. For a service that gates deployment pipelines at thousands of companies, that is a meaningful number — and Actions is consistently the weakest component in that list.

Developers on Hacker News have been raising the same point for weeks, arguing GitHub has felt less stable through 2026 than in previous years. That is a perception, not a measurement, and GitHub has not commented on it. But it is a perception held by people who have used the platform for over a decade.

The lesson worth taking

Most teams treat GitHub as infrastructure, not as a vendor. It is a vendor.

Two questions worth answering after today, while the frustration is fresh:

If GitHub were unavailable for a full working day, could you still deploy to production?
Does anything in your build pipeline fetch from GitHub at runtime rather than from a cache or internal mirror?

If either answer is uncomfortable, mirroring critical dependencies internally costs far less than an unplanned day of blocked releases.

Key Takeaways

GitHub is down worldwide as of August 17, 2026, and the incident remains unresolved at the time of writing.
Five services are in major outage: API Requests, Issues, Pull Requests, Actions and Copilot.
Error rates are approximately 20% for web and API, and 50% for archive downloads and raw content.
Copilot is down, but its underlying AI model providers remain operational — the fault is in GitHub’s service layer.
Downdetector recorded 3,000+ reports, with 849 from India.
GitHub has not published a root cause. Claims of an Azure link are unconfirmed.
Local git work is unaffected. Your commit history is safe.

FAQs

Is GitHub down right now?

Yes. As of 21:12 IST on August 17, 2026, GitHub’s status page lists API Requests, Issues, Pull Requests, Actions and Copilot as being in a major outage. GitHub is applying mitigations.

Why is GitHub Copilot down?

Copilot is affected as part of the wider GitHub incident. GitHub’s status page shows Copilot in major outage while Copilot AI model providers remain operational, indicating the failure is within GitHub’s own infrastructure rather than the AI models.

Is my code safe during a GitHub outage?

Yes. Git is a distributed version control system, so your full repository history exists in every local clone. An outage prevents access to GitHub’s servers but does not affect code already committed locally.

Can I still use git while GitHub is down?

Yes. Local operations — commit, branch, merge, log, rebase — work normally. Only operations that contact GitHub’s servers, such as push, pull, fetch and clone, will fail or be unreliable.

What caused the GitHub outage on August 17, 2026?

GitHub has not disclosed the cause. The investigation is ongoing and no root cause analysis has been published.

How long will the GitHub outage last?

GitHub has not given an estimated resolution time. The company confirmed the incident at 9:40 AM EDT and began mitigations at 11:42 AM EDT, with error rates unchanged since.

Where can I check GitHub’s official status?

githubstatus.com is the authoritative source. Third-party monitors reported conflicting states during this incident.

Also Read: 100-Foot Tricolour Installed at Doda Stadium for 80th Independence Day (15 Augest 2026): Why It Matters in Chenab Valley Jammu Kashmir