NSANet, the National Security Agency's classified internal network, went down on Tuesday, and by that night the agency had closed its Fort Meade campus to all but essential staff. Director of National Intelligence Jay Clayton blamed the NSA network outage on patches for PHP vulnerabilities discovered last month, saying some systems did not come back up as expected. Any business rolling out the same PHP fixes faces the same gap between installing a patch and getting the service healthy again.
Key takeaways
- NSANet, the NSA's classified internal network, went down on Tuesday; Fort Meade closed to all but essential staff that night, and restoration was still incomplete in the latest reporting.
- DNI Jay Clayton said patches for PHP vulnerabilities found last month left some systems that did not come back up as expected; CNN's sources saw no evidence of a cyberattack at the time.
- Officials said communications with the traveling President and Five Eyes intelligence sharing were unaffected.
- No CVE or PHP version has been named. For businesses, the lesson is the restart: save logs first, verify fixes against distribution advisories, restart serving processes and test real pages.
What happened to the NSA network on Tuesday?
NSANet, the network NSA employees use to share classified information internally, went down on Tuesday, according to six people familiar with the matter who spoke to CNN. In the latest reporting, restoration was ongoing but incomplete, with no date given for full service.
Staff learn whether to report to Fort Meade from a voicemail inbox, and it changed three times in one day:
- Tuesday morning, code yellow. The second-highest of three main levels: non-emergency staff came in staggered shifts, emergency personnel as usual. The New York Times first reported this detail.
- Tuesday evening, code green. Employees were told to report as scheduled.
- Later Tuesday night, code orange. The campus closed to most employees until further notice.
Nobody has explained the reversal. Read the codes as staffing orders, not a network health check: green did not mean NSANet had recovered, and orange shows the agency wanted people off campus, not what was broken.
Was the NSA network outage a cyberattack or a failed patch?
Every official account points to maintenance gone wrong. CNN’s sources saw no evidence of a cyberattack “at this point,” and the NSA described “temporary degradation during scheduled routine maintenance,” saying mission-critical capabilities “remain fully operational” thanks to multiple layers of redundancy.
Jay Clayton went further on Fox Business, tying the outage to patching major PHP vulnerabilities discovered last month. The patching “caused some of these systems to not come back up when you patch them as expected,” he said, adding that none of the recent tech outages, the NSA’s included, shows signs of an attack.
According to the report, the flaws could let attackers break into websites and read memory holding passwords and other secrets. AI tools helped researchers exploit them, setting off a patching scramble worldwide, and the NSA itself urged organizations to patch. None of that says anyone attacked the NSA.
No CVE, affected PHP version or fixed build has been named. Until one is, check the security entries in the PHP changelog on php.net and your Linux distribution’s security tracker (Debian, Ubuntu and Red Hat all publish one), which lists the fixed package build for your release.
Did the outage cut off the President or Five Eyes?
No, according to officials who spoke to CNN. A current and a former US official said urgent intelligence would still reach President Donald Trump on Air Force One. He was in the Philippines, heading to China the next day for talks with Xi Jinping and Vladimir Putin. A source from one Five Eyes country and a former US intelligence official said sharing among the US, UK, Canada, Australia and New Zealand was unaffected.
A former senior NSA official called NSANet “more of a network of networks,” so a full outage is unlikely: analysis and reporting could slow, but the highest-priority work would find another route. The verdict: “alarming, but likely not crippling.” In 2000, a technology failure took Fort Meade computers down for nearly three days. With restoration still incomplete, any comparison is premature.
Why can a PHP patch keep systems from coming back up?
Installing a patch and returning a service to health are separate events, and outages happen in between. After the PHP package updates, running PHP-FPM pools, Apache workers, queue consumers and containers keep executing the old code until something restarts them.
The restart is where things break: an extension compiled for the old build refuses to load, a config file has drifted, a dependency is missing, or a batch of hosts reboots at once and runs out of memory. Clayton did not say which of these, if any, hit the NSA. Our earlier piece on the PHP patch that still needs a restart covered the opposite failure, where the old, vulnerable code keeps running.
The report says PHP powers about 70 percent of the world’s websites, so emergency patch windows are running everywhere at once. From the outside, a failed restart and an intrusion look alike. Only logs settle it.
What should an SMB check after a PHP security patch?
Check the path that serves requests, not the version string, and save evidence before you start rebooting.
- Save evidence first. Copy service journals, package-manager history, config diffs and health-check output. The mistake we see most after a bad patch night is the third or fourth reboot that wipes the only logs explaining the first failure.
- Verify the fix, not the version. Distributions backport security fixes under the old upstream version number, so the PHP version command proves nothing. Compare the installed package against your distribution’s advisory. If your branch no longer gets security updates, plan and test an upgrade.
- Restart what serves traffic. PHP-FPM, the web server and long-running workers. Rebuild container images instead of patching inside a running container.
- Test real pages. Hit a page that queries the database, submit a form, run a scheduled job.
- Stage the rollout. Patch one host, watch it, then do the rest, with a rollback snapshot you have actually tested.
- Tell staff what is down. One line: what is affected, what works, when the next update lands.
Those records give you a timeline (deployment job, package change, restart output, first error) that tells your board or insurer whether this was maintenance or an attacker.
If the app faced the internet and you patched a memory-disclosure flaw late, treat it as a possible incident. Patching stops new leaks; it does not revoke old ones. Review access logs, then rotate credentials, API keys and session secrets in order of risk. Odd requests or unknown accounts call for a compromise assessment, and a standing vulnerability management program keeps an unsupported PHP branch from surprising you next time.
What is still unknown about the NSA network outage?
Officials have not said which PHP flaws the NSA was patching, which systems failed to return, whether the patch or the restart broke them, or when NSANet will be fully restored. The lack of attack evidence reflects what investigators knew on Tuesday, not a final verdict.
FAQ
Was the NSA hacked?
Officials say nothing points that way so far. CNN's sources saw no sign of a cyberattack, and DNI Jay Clayton blamed PHP patches that left some systems that did not come back up as expected.
Why did the NSA close Fort Meade?
After NSANet went down, staff status moved from code yellow to code green and then to code orange late Tuesday, limiting the campus to essential personnel. The agency has not explained the reversal.
Were the FAA, Microsoft and Chase outages connected to the NSA?
Nothing in the reporting links them. The FAA traced Friday's ground stop to a single router in Arkansas, the report sees no sign the NSA, FAA or Microsoft outages were cyberattacks, and it gives no cause for the Chase online-banking problems.
What should I do if my site stays down after a PHP patch?
Save logs and config diffs first, then check PHP-FPM, web server and worker restarts, failed extensions and container images. Roll back to a tested snapshot if needed, and rotate secrets if the app was exposed before patching.
If a patch night has ever turned into a lost day, test your rollback before the next one. We can help you build a recovery plan that gets services back without guessing.
Tell us what you're running
Scoping is free. We reply within one business day, and under 30 minutes for active incidents.
