GitHub Outage Map

Community Discussion

Tips? Frustrations? Share them here. Useful comments include a description of the problem, city and postal code.

Beware of "support numbers" or "recovery" accounts that might be posted below. Make sure to report and downvote those comments. Avoid posting your personal information.

GitHub Issues Reports

Latest outage, problems and issue reports in social media:

• 
  Mihai Maruseac (@mihaimaruseac) reported 3 minutes ago

  I know a lot of great folks at GitHub and I'm saddened to see all these issues. But replying to this incident that impacted the integrity of repositories and broke the mental model of *** operations with a "roughly 0.07% were affected" is not right. I fear for GitHub's future.

• 
  Vitalii Khomenko (@VitaliiKhomenk1) reported 8 minutes ago

  @_rajathhr @NathanMcNulty You clearly don’t use Claude Sonnet in GitHub Copilot—it hasn’t been working there for over two years. You send a request, it starts thinking, and then you just get an error—that’s it, tokens wasted. I’ve written about this a thousand times.

• 
  Wassollichhier (@wassollichhier) reported 11 minutes ago

  @openclaw @cherry_mx_reds Pls fix the Confy UI integration. There are multiple issues already on github

• 
  Vulnerable U (@vuln_u) reported 14 minutes ago

  Unlike basic data theft, it encrypts and sends data to an external server, and if that fails, it pivots to GitHub creating repos, abusing tokens, and even pulling secrets from CI environments. If version 2026.4.0 is installed, assume everything on that machine is compromised.

• 
  Felix Allistar (@FelixAllistar) reported 16 minutes ago

  @kr0der subagents + github. OAI models already do a mini review at the end of large turns, and i've found a clean context window is less likely to say that its fine now, because it already said it was fine last turn. issue was really bad in -codex models, and 5.5 feels like a hybrid. subagents=new chat, it uses the same thing as /review but lets it stay pinned to your parent megathread.

• 
  Sarath 👨‍💻 (@sarat) reported 17 minutes ago

  @GergelyOrosz Moreover - they should cut down stuff like azure devops and integrate with GitHub. It’s redundant, substandard and none is building for ADO

• 
  junior85 (@junior85) reported 21 minutes ago

  @thsottiaux You need to take a look at the issues on Github. Windows app is fecked.

• 
  Rachel Blum (@groby) reported 23 minutes ago

  @hkarthik @satyanadella You mean after he fixed Windows? And their office apps? Snark aside, I don't think Github is a "Satya" shaped problem per se - the culture that leads to these things is, though. Except... it smells like the culture reverted to pre-Satya norms.

• 
  Riyaz Saiyad (@Rj_Saiyad) reported 25 minutes ago

  @github I’ve been overcharged for Copilot Pro (Ticket #4299654). You charged me $10 instead of the promised $5.11, counting free trial requests as paid! 😡 No response for 7-10 days while my money is gone and I have no access. Fix this ASAP! #GitHub #Copilot

• 
  24AIGlobal (@24AInor) reported 26 minutes ago

  Gas Town allegedly running inference on your own API credits without telling you. GitHub issue with 248 HN votes, still no official response. open source doesn't mean transparent. worth auditing what your tools actually send out.

• 
  Got Eem (@HaGoshem) reported 30 minutes ago

  @Longsword44 i also want to mention that this **** breaks constantly but the devs fix it like immediately. broke sometime this morning but the devs fixed it 3 hours ago on github

• 
  Julian Goldie SEO (@JulianGoldieSEO) reported 30 minutes ago

  There's a free open-source model that copies Anthropic's biggest secret. It is called OpenMythos. A guy named Kai Gomez built it from scratch on GitHub. Big AI models cost a fortune because they are massive. This new model is different. It loops its own brain to think deeper about hard problems. You get smarter answers without paying for a giant data center. This means you can run powerful AI right on your own machine. Your data stays private. Your API bills drop to zero. You can finally automate your emails and support tickets for free. If you want the exact steps to set this up for your business, join the AI Profit Boardroom.

• 
  Ronan Berder (@hunvreus) reported 30 minutes ago

  @badlogicgames What about greenfield work, or new features. I'm assuming you don't do that off of GitHub issues. Just talk back and forth with the AI to build a plan in a markdown file and then pull the trigger once aligned?

• 
  Atarus (@Atarussecurity) reported 32 minutes ago

  A critical SSRF vulnerability in LMDeploy, the open-source toolkit used to deploy large language models, was disclosed on GitHub last week. Twelve hours and thirty-one minutes later, an attacker was already in the honeypot. Eight-minute session. Ten requests across three phases. They confirmed the SSRF with an out-of-band DNS callback, enumerated the API surface, and ran a full internal port scan reaching AWS Instance Metadata Service, Redis, MySQL, and an admin interface behind the model server. No proof-of-concept code existed at the time of the attack. The attacker built theirs from the advisory text alone. This is the third major AI infrastructure exploitation we have analyzed in five days. The patch-to-exploit window is now measured in hours, regardless of install base size. Self-hosted AI infrastructure is a first-class attack surface. Treat it accordingly.

• 
  Filip (@filip_a__) reported 33 minutes ago

  @nbrempel @scaling01 terrible.. now u get 40 gpt 5.5 messages per month on 10$ plan and 200 on 40$ plan and thats if u dont get rate limited to hell, oai has about 450 requests per month on 20$ plan if u only do 15 requests per day with less rate limiting. idk what they are smoking at github recently