By Ayusha Oli, Cybersecurity Operations, Jutsu
Okay so I need to vent for a second before I get into this properly, because I have now read approximately one billion breach disclosures this week and I am starting to feel like a doctor who only ever sees patients after they have already ignored every symptom for six months.
Here is the last 48 hours, presented the way it actually felt to read it, which is less “fascinating threat intelligence” and more “oh no, not again.”
First up, MikroTik. CERT Polska put out a warning that attackers are getting full administrative control over MikroTik routers with zero authentication, just by hitting SSH ports that were left open to the entire internet. No password guessing, no cleverness, nothing. The digital equivalent of finding out your front door was not just unlocked, it was propped open with a brick. MikroTik has patches out. If you have one of these boxes anywhere in your network, please, I am begging you, go check it before I have to write about you next.
Then JetBrains, which honestly is the one that made me put my head down on my desk for a minute. Someone broke into their Cadence cloud service through an unpatched TeamCity flaw and walked out with AWS credentials. JetBrains’ official advice to every customer is basically “assume everything you built using our tool for the last while might be compromised, please rotate literally everything.” A build tool. The thing you trust to just quietly build your code in the background is now on the list of things you have to interrogate. Great. Cool. Love that for all of us.
Meanwhile, over in e-commerce land, a researcher found an unpatched, unauthenticated remote code execution bug in Magento and Adobe Commerce, cheerfully named StyleSmuggler, and published it early because stores were being actively backdoored while the report was still being written. Adobe had not even put out an advisory yet. So if you run a Magento store right now, that is your unscheduled Tuesday.
And then, as a little cherry on top, OpenAI casually confirmed that GPT-6 Astra scored a perfect 100 percent on their exploit-finding benchmark. One hundred percent. Not “pretty good at security.” Perfect. The model that can, apparently, find and weaponize the next MikroTik, the next JetBrains, the next StyleSmuggler, all on its own, while I am over here manually checking router configs.
I want to be annoyed that none of this is new information, structurally speaking. It is the same lesson every single week: something we trusted to just quietly do its job (a router, a build pipeline, a shopping cart, a login form) turned out to be the whole attack surface the entire time. But I think that is exactly why it is exhausting. It is not that the attacks are getting smarter. It is that we keep leaving the same kinds of doors open and being shocked, shocked, when someone walks through one.
Anyway. This is basically my entire job now, reading things like this at eleven at night and then building systems that are supposed to catch this stuff before it becomes a headline instead of after. Some weeks that feels manageable. This week it mostly felt like everyone’s homework was due at once.
If your SOC team also feels like it is drowning in exactly this kind of nonstop noise, that is genuinely what we are building AgentSOC for at Jutsu. Come commiserate with us.