Cybersecurity is now a race for speed
Attackers no longer give you days to react — they give you hours. According to Rapid7’s Q3 2026 threat report, high- and critical-severity vulnerabilities are exploited 3 days after disclosure on average, and 71% are hit within the first week. And the harshest number comes from Mandiant, in its M-Trends 2026 report: the average time to exploit a vulnerability has dropped to an estimated -7 days. Minus seven: they exploit it before the patch exists.
The week of November 10–18 left the clearest proof: one compromised vendor, and suddenly 178 organizations. Cybersecurity stopped being a static defense and became a race for speed.
The attack of the week: one vendor, 178 companies
On November 17, a supply-chain attack was reported that started with Optimizely, an experimentation platform that hosts scripts loaded by thousands of websites. According to Newsweek, the attackers injected malicious JavaScript into those scripts and reached 188 projects across 178 organizations: BBC, CNN, Zoom, HSBC, Salesforce, GitLab, IKEA, Equifax, Air France-KLM, Ledger, Peacock, Fox News, and Alaska Airlines, among others.
The code stole card data, changed cryptocurrency wallet addresses, captured cookies and screens, and redirected older iPhones and iPads to an exploit kit. Scott Helme, a researcher at Report URI, estimated about 34 million page views loaded the malicious code (a range of 6 to 150 million). His line sums it up: “Compromise one trusted vendor, and its code runs on every customer’s website at once.”
The window closed: from days to hours, and to minus seven
The numbers from this same season draw the acceleration with precision. The same report adds that handoff from initial access to a secondary actor dropped from more than 8 hours in 2022 to 22 seconds in 2025. And median dwell time — the time an attacker lives inside your network unseen — now stands at 14 days globally, according to the same report.
Rapid7 measures the same acceleration from the other side, and every figure is a clock. The average time to exploit a high- or critical-severity vulnerability is 3 days, and 71% are attacked within the first week; with Mandiant the clock runs to minus 7 days, before the patch exists. 79% are exploited remotely, with no credentials and no click from anyone: no person opens the door, a script does it in seconds. 80% land in the first two weeks and 90% in the first month, so whoever patches on the last Friday of the month arrives after that month has already been exploited.
Why patching is no longer enough
For years, defense boiled down to a ritual: patch at the end of the month, audit once a year. It still helps, but it no longer protects. If the vulnerability is exploited before the patch exists (Mandiant’s -7 days), the “discover → patch → rest” cycle always arrives late. And if the attacker jumps from one system to another in 22 seconds, a team that responds “as soon as the office opens Monday” is not playing the same game.
The only real defense that survives this speed is to run at the same speed — or faster. That means three concrete things:
- Continuous monitoring. Constant watch over what comes in and goes out, not a one-off scan. A 14-day dwell time exists because nobody is watching in real time.
- Early detection. Seeing suspicious movement the moment it happens, not in the end-of-month report.
- Automated response. Your own team of defensive bots and agents that isolates a compromised access, blocks a route, or revokes a session in seconds, while a person decides and supervises.
In other words: if the attacker uses automation to exploit at scale, you need automation to defend at scale. An annual checkup does not compete with an adversary that acts in hours.
What to review this week
You don’t need a bank-sized budget to start closing the speed gap. If your company has between 5 and 30 people, review this today:
- MFA everywhere. Two-step authentication on email, banking, and admin panels closes most credential-theft access, and it takes an afternoon to turn on.
- Patches on anything internet-facing. What faces the internet gets updated first and with priority, not in the monthly batch: there the window is measured in hours.
- Who responds, and how fast. Define who gets the alert and what they do in the first 5 minutes. Speed is trained, not improvised.
- Tested backup. If everything else fails, this is what gets you out of the incident: a backup you have never restored is not a backup.
- Inventory of third-party vendors and scripts. Which third parties load code on your site or in your systems? Write them down, one by one. They are your side doors.
Defense is built before the attack
Here is the real shift: cybersecurity is no longer something you buy once there is an incident. It has to be built, tested, and running before. At today’s speed, nobody improvises a response team at three in the morning on the day it happens.
That is why “having cybersecurity” today means having your own team of defenders —people and AI agents— watching, detecting, and acting in seconds across your network and systems. And it means having it ready before the attack. If you build it while you are being attacked, you are already late: there is no time to learn. Either you run with your team from before, or you are history.
At PG Tech we are re-arming our cybersecurity offering exactly that way: the defense team is prepared, tested, and trained before it is needed. Continuous monitoring and agents that respond in seconds, with a person supervising and deciding.
