On Tuesday evening, 20 October, the identifiers for two critical flaws in PHP leaked online while they were still under embargo. There was no patch.
On Wednesday afternoon the Australian Cyber Security Centre issued a Critical Alert: Act Now. That’s the highest rating it uses. Its first instruction wasn’t to patch. It was to identify where PHP is running across your environment. [1]
By Wednesday night the national cyber agency’s own website was offline. So were several other federal government sites. The minister’s office said some were “temporarily unavailable as work is undertaken to mitigate risks”. [2]
Picture who was at the other end of that alert.
The head of security at a mid-sized super fund, who’d spent the morning on a CPS 230 remediation item and the afternoon finding out how many PHP servers the fund actually had. The one-person security function at a regional water utility, ringing three suppliers at 9pm to ask whether their portals ran PHP, and getting voicemail. The IT manager at a forty-person accounting practice, four months into AML obligations, who had never heard of PHP and was now responsible for it.
None of them slept much last week. Most of them won’t get a thank you.
This post is about them.
Nobody could have seen this coming
Let’s be clear about this first, because blame tends to land quickly after a week like that.
The two flaws had been sitting in PHP for roughly fourteen years. [3] They weren’t in an application anyone built. They were in the engine underneath millions of applications. And they became public the wrong way, through a leak, before anyone had a fix to give you.
Then the attackers moved faster than the patch. Amazon logged an exploitation attempt on a honeypot on 17 October, three days before the leak and four days before the fix. Cloudflare saw thousands of attempts in the ten hours before the announcement, tens of thousands in the forty-five minutes after it, and hundreds of thousands within two hours. [4]
No risk register had this on it. No audit would’ve found it. No framework, maturity level or control set would’ve prevented it. If your security team spent last week scrambling, it wasn’t because they missed something. It was because the whole world got the same news on the same night, and they were the ones who had to act on it.
Most of what’s been written about last week is about how the flaws were found. That’s a real story. It isn’t this one.
If your team spent last week scrambling, it wasn’t because they missed something. The whole world got the same news on the same night.
We’ve been here before
If last week felt familiar to anyone in your security team, it’s because they’ve lived it.
In December 2021 it was Log4j. A flaw in a logging library buried inside software nobody knew they had. The head of the US cyber agency called it “one of the most serious I’ve seen in my entire career, if not the most serious”. [5] It landed two weeks before Christmas, and a lot of security people spent that Christmas hunting for it.
The comparisons were made within days. Tenable rated last week’s flaws among the most consequential since Log4Shell. [3] One researcher put it at “like 50% the way towards Log4J”: a worse direct hit, but a shorter tail. [4] A UNSW professor called it “the biggest cyber security emergency in a decade. Again.” [3]
That word “again” is the one to notice.
After Log4j a US review board concluded the flaw had become “endemic”, with unpatched copies likely to remain in systems for the next decade, if not longer. [6] Organisations learned a lot about software inventories that December. What most of them didn’t change was how many people carry the weight when one of these lands. It’s still the same small group. And they were already full.
They were already full before Tuesday
Last week we counted what 2026 has added to an Australian organisation’s cyber obligations. Depending on your industry it’s somewhere between four and eleven new obligations, and almost none of them replaced anything. [7] They went on top of the pile, mostly on the headcount approved last December.
That’s the team that got the alert on Wednesday afternoon.
The research on how they were coping before last week isn’t comfortable reading. In Proofpoint’s 2025 survey of Australian CISOs, 68 per cent said they’d experienced or witnessed burnout in the past year, and 75 per cent said they faced excessive expectations. [8] In a 2025 global study of CISOs, nearly half said burnout had already weakened their ability to prepare for a breach. [9] Among cyber security professionals surveyed by Cybermindz, 60.5 per cent said they felt “used up” at the end of the workday daily or weekly. [10]
And here’s the finding boards should sit up for. In that same global study, 44 per cent of CISOs named board and executive expectations as their biggest source of stress. Only 33 per cent named external threats. [9]
So the biggest pressure on these people isn’t the attackers. It’s us.
Gartner saw this coming three years ago. In 2023 it predicted nearly half of cyber security leaders would change jobs by 2025, with a quarter leaving for different roles entirely. Its explanation was blunt: “Burnout and voluntary attrition are outcomes of poor organizational culture.” [11] Not outcomes of the threat landscape. Of how organisations treat the people defending them.
44 per cent of CISOs name board and executive expectations as their biggest stressor. Only 33 per cent name the attackers.
So start by saying thank you
This is the cheapest thing on the list, and it’s the one most often skipped.
Your security team, your IT operations people, your developers and the supplier staff who picked up the phone at 9pm did something hard last week. They found software they didn’t know they had. They patched it under pressure. They took services down when they had to and brought them back up. They checked logs for signs someone got in before the door was shut. And they did it on top of a week that was already full.
Thank them specifically. Not a line in the CEO’s all-staff email. Have the chair or the CEO say it to them directly, name what they did, and put it in the board minutes. People notice whether the board noticed.
Then give them some time back. If people worked nights last week, they shouldn’t walk into a normal deadline this week.
What last week actually tested
Here’s the useful lesson, and it isn’t about preventing the next one. You can’t. Greg Brockman, president of one of the companies whose tools helped find these flaws, said it plainly: “we expect to find many more such issues over upcoming months”. [3]
What you can control is how hard the week is for the people who carry it. Last week tested three things, and none of them is technical. Did you know what you were running? Could the right person make the call? And how many people carried it?
Organisations that had good answers to those three had a long week. Organisations that didn’t had a terrible one, and it landed on the same few people.
Three things every ExCo and board can do
One. Fund the inventory before you need it.
The ACSC’s first instruction was to find PHP across your environment. [1] As one industry director put it, “If you don’t know what software you’re running, you can’t know whether you’re vulnerable.” [12]
If that took your team three days, those were three days of people working nights doing archaeology. Ringing suppliers. Logging into servers nobody owns. Searching old project folders. That isn’t response work. It’s catch-up work, and it’s the most exhausting kind, because it feels like it should already have been done.
A current inventory of what you run, including what your critical suppliers run on your behalf, is a security control. It’s also a wellbeing control. It turns a three-day hunt into a three-hour query.
The board question: if ASD told us tomorrow to find every instance of one piece of software, how long would it take, and who would be doing it?
Two. Agree in advance who can pull the plug.
Some of the government sites went offline last week because, for a few hours, the safest thing to do was stop serving. [2] Somebody had to make that decision, at night, with incomplete information.
In a lot of organisations the person who knows a service needs to come down isn’t the person allowed to take it down. So they spend the night chasing approvals instead of fixing things. They write the case, wait for a reply, escalate, wait again. And if they make the call without approval, they carry it personally.
Fix that before the next one. Decide now which internet-facing services the security lead can take offline on a critical alert without asking first. Agree the business impact you’ll accept to do it. Make the executive team informed, not consulted. Then rehearse it once, so it isn’t the first time at 11pm.
The board question: on a night like Wednesday, who in this organisation can take a service down, and does that person know they can?
Three. Plan for the surge and the recovery, not just the response.
Most incident plans describe the response. Very few describe who does the work when the normal team is already at capacity, or what happens to them afterwards.
So plan for both. Have surge capacity contracted before you need it: an incident response retainer, a managed service provider with agreed hours, or a fractional security lead who already knows your environment. Make sure the on-call roster isn’t the same two people every time. And decide in advance what gets dropped when an emergency lands. If the team spends a week on a critical alert, something on this year’s list of four to eleven obligations moves. Say which one, out loud, so nobody tries to do both.
Then build in the recovery. Time off after a week like that isn’t a perk. It’s how you still have a security team for the next one.
The board question changes too. Not “are we patched?” Ask “what did this cost the team, and what did we drop to do it?”
Don’t ask “are we patched?” Ask “what did this cost the team, and what did we drop to do it?”
Where BayRisk lands on this
We don’t think last week was a failure by security teams. The opposite. Most of them did something difficult, quickly, with too little information and not enough people. A lot of organisations are fine today because somebody gave up their week.
But we do think it showed how thin the margin is. The same small group carries the new obligations, the regulator letters, the board reporting and the emergencies. When one of these lands, there’s nothing left to absorb it. That’s not a resilience position. It’s luck, and it runs out.
None of the three things above needs a new technology budget. An inventory, a decision right and a surge plan are governance decisions. They sit with ExCo and the board, not with the security team. That’s where we think the responsibility for how hard the next week is belongs.
This is the work RiskOps does the other fifty-one weeks of the year. Keeping the picture of what you run current. Making sure decision rights are written down and tested. Being the extra senior pair of hands when the normal team is full. So that when the next one lands, and it will, it’s a hard week rather than the worst week of somebody’s career.
If your team had that week, start by telling them you noticed. Then, if you’d like help with the three things above, that’s a short conversation: hello@bayrisk.com.au.
References
- Cyber Daily, “Critical alert: Aussie cyber agency warns businesses to monitor PHP vulnerabilities”, 21 October 2026. https://www.cyberdaily.au/security/14334-critical-alert-aussie-cyber-agency-warns-businesses-to-monitor-php-vulnerabilities
- Australian Community Media, “Government cybersecurity website hit in global flaw”, 22 October 2026. https://www.newcastleherald.com.au/story/9373072/government-cybersecurity-website-hit-in-global-flaw/
- iTWire, “PHP2Shell: the 14-year-old bugs that pulled cyber.gov.au and government sites worldwide offline”, 23 October 2026. https://itwire.com/business-it-news/security/php2shell-the-14-year-old-bugs-that-pulled-cyber-gov-au-and-government-sites-worldwide-offline
- SecurityWeek, “PHP2Shell exploitation started days before patches”, 22 October 2026. https://www.securityweek.com/php2shell-exploitation-started-days-before-patches/
- Jen Easterly, then Director of the US Cybersecurity and Infrastructure Security Agency, on Log4Shell, December 2021, as recorded at Wikipedia, “Log4Shell”. https://en.wikipedia.org/wiki/Log4Shell
- The Record, “First Cyber Safety Review Board report finds Log4j has become an ‘endemic vulnerability'”, 14 July 2022. https://therecord.media/first-cyber-safety-review-board-report-finds-log4j-has-become-an-endemic-vulnerability/
- BayRisk, “Count what 2026 added to your cyber obligations. Then look at whose name is on it.”, October 2026. https://bayrisk.com.au/blog/
- IT Brief Australia, “Australian CISOs under strain from AI risks, burnout and attacks”, reporting Proofpoint’s 2025 Voice of the CISO report, 27 August 2025. https://itbrief.com.au/story/australian-cisos-under-strain-from-ai-risks-burnout-attacks
- Help Net Security, “CISOs are cracking under pressure”, reporting Nagomi Security’s 2025 CISO Pressure Index, 11 November 2025. https://www.helpnetsecurity.com/2025/11/11/stress-ciso-burnout-crisis/
- SecurityBrief Australia, “Cybermindz warns mental health absences hit cyber staff”, 26 March 2026. https://securitybrief.com.au/story/cybermindz-warns-mental-health-absences-hit-cyber-staff
- Gartner, “Gartner predicts nearly half of cybersecurity leaders will change jobs by 2025”, press release, 22 February 2023. https://www.gartner.com/en/newsroom/press-releases/2023-02-22-gartner-predicts-nearly-half-of-cybersecurity-leaders-will-change-jobs-by-2025
- Australian Computer Society, Information Age, “Govt sites go down after AI finds critical PHP vulnerabilities”, 22 October 2026. https://ia.acs.org.au/article/2026/govt-sites-go-down-after-ai-finds–critical–php-vulnerabilities.html