Explore our new Cybersecurity Terms Glossary, explained by our experts.
Cybersecurity Terms Glossary
On 19 July 2024 a faulty content update to CrowdStrike Falcon crashed roughly 8.5 million Windows machines worldwide. It was not a cyberattack, but a change management failure at global scale.

What Happened in the CrowdStrike Outage?
Shortly after 04:00 UTC on 19 July 2024, Windows machines running CrowdStrike Falcon began crashing into the blue screen of death and rebooting into a loop. Airlines grounded aircraft, broadcasters went off air, payment terminals stopped, and in the UK a number of GP practices lost access to clinical systems. Microsoft later estimated that around 8.5 million Windows devices were affected. No attacker was involved. A security product deployed to protect those machines took them down.
Falcon receives two kinds of update: sensor versions, which customers can stage and delay, and Rapid Response Content, which ships behavioural detection data and historically bypassed that staging because threats move quickly. On 19 July CrowdStrike published a Rapid Response Content file, Channel File 291, containing a template instance with more input fields than the sensor’s Content Interpreter expected. The mismatch produced an out-of-bounds read in kernel mode. Because Falcon runs as a kernel driver, an invalid memory access is not a crashed application; it is a crashed operating system.
CrowdStrike’s preliminary post incident review, published on 24 July 2024, and the fuller root cause analysis that followed on 6 August identified the gap plainly: a content validator had passed the file despite the parameter count mismatch, and there was no staged rollout or canary deployment for this content class. A single publication reached every host subscribed to that channel within minutes.
The fix itself was trivial — delete the offending file — but the machines could not boot far enough to receive it. Administrators had to start each host in safe mode or the recovery environment and remove C-00000291*.sys by hand. On BitLocker-protected machines this required the recovery key, which for many organisations lived in a management console that was itself running on affected servers. Estates with remote staff, kiosks, or point-of-sale terminals had no practical route other than physically visiting each device. Delta Air Lines publicly claimed losses of around $500 million.
Boards asked the wrong question afterwards. The useful one is not “how do we avoid CrowdStrike” but “which suppliers can push code into our kernel, or our build pipeline, or our browsers, without our approval, and what happens if one of them is wrong?” Endpoint agents, EDR sensors, antivirus engines, browser extensions, managed device policies, and continuous deployment tooling all share that property.
For UK financial entities the incident landed in the middle of the operational resilience regime: firms must set impact tolerances for important business services and be able to stay within them during severe but plausible disruption. A vendor pushing a bad kernel driver to every endpoint simultaneously is exactly the scenario those rules contemplate, and it is now a standard scenario in third-party dependency testing. EU-authorised entities face the same question under DORA, which treats concentration in critical ICT third-party providers as a supervised risk rather than a commercial choice.
At Privia Security, we use this incident as a scenario in cyber exercise services precisely because it is uncomfortable: there is no adversary to blame and no patch to apply, only decisions made long before the day.
No. It was a defective content update to the Falcon sensor. A configuration file contained more input fields than the sensor’s interpreter expected, causing an out-of-bounds memory read in a kernel driver and crashing Windows. No adversary was involved, although criminals did exploit the confusion afterwards with fake recovery tools and phishing that impersonated support channels.
Affected machines crashed during boot, so they never reached a state where they could contact the management network to receive a corrected file. Remediation required booting into safe mode or the recovery environment and deleting the offending driver file manually, which on BitLocker-protected systems also required the recovery key to be available.
Yes. Any product with a kernel-mode component and an auto-update channel carries the same structural risk, as do browser extensions, device management agents, and deployment pipelines. The mitigation is not avoiding security software but knowing which suppliers can change production without approval, using whatever staged rollout controls they offer, and testing recovery.
Three things, in order. Configure staged or delayed updates wherever a vendor supports them. Store disk encryption recovery keys somewhere that does not depend on the systems being recovered. Then rehearse mass manual remediation to find out how many devices your team can realistically restore per day, including machines held by remote staff.
You May Be Interested In These
We use a few cookies that are needed to make this website work, and they are always on. We would also like to set analytics cookies, to see which pages are useful; functional cookies, so that our live chat works; and advertising cookies, to measure our campaigns. We will not set any of those until you say yes, and you can change your mind at any time from the Cookie Settings link in the footer.
Choose the cookies you are happy for us to use. Strictly necessary cookies are always on; everything else is off until you turn it on.
These cookies keep the website working: they hold your security checks, your form submissions and the choice you make here. The site cannot run without them, so they are always on.
These let our live chat open and remember your conversation, and let the newsletter sign-up work. Switch them off and the chat window will not appear.
These tell us which pages are read and how people find us, so that we can improve the site. The reports are about groups of visitors, never about you personally.
These let us measure our advertising and show you our adverts on other websites, such as LinkedIn, Google and Meta. Switch them off and you will still see adverts, but they will be less relevant.