I wonder (and have no idea) whether the increased reports of services being down is a case of the Baader–Meinhof phenomenon, or a actual increase in failure rate.
If I was asked to armchair-pattern-match, I'd say that there's a combination of two things: (a) a wider fear-based environment where rapid modernization is being pushed heavily, and (b) an increase in tools that give false confidence on change management.
AI is tuned to speak confidently and ship independently; it is rare that it will suggest that one go back to a broad survey of stakeholders to ensure no Chesterton's Fence is being torn down (https://fs.blog/chestertons-fence/). "I've read the documentation and logs, and nothing suggests this system needs these legacy capabilities/components/capacity" is something we hear all day, and it is a siren's song if there ever was one.
Couple that with a general (and valid!) consensus that modernization of legacy systems must be accelerated for a variety of reasons - to mitigate this new and continuing "Black October" of hyper-agile agent-driven cyberattacks (whether the network is public or private); to have more flexible infrastructure for new features to stay ahead; to allow human teams to show their efficacy in a world where technology jobs are increasingly at risk - and you have a perfect storm where change management is no longer what it once was.
That's my impression; a total of zero of the popular recent ones have been related to "AI" or Cyber attacks, but every comment in every thread, talks about those two as if they were the underlying cause.
The FAA's recent outage for example, at its core, seems like people trying to rush needed upgrades that were slow for a specific reason (e.g. reducing the testing window for 90 to 30-days).
The Microsoft outage seems like three different things linked together (and poor "full failure" recovery prep, lengthening the outage).
The bank "outage" may not have even existed, 2/3 of the banks claimed they didn't have an outage at all, and the other said it was a limited technical issue. The articles were based on DownDetector spikes.
You need to be more skeptical of what organizations say when they have a lot to lose. Nobody takes responsibility for situations unless there's way to offload some or all of the blame. They have every incentive to lie about the situation to make it go away. Politics often plays the biggest role in these situations.
It's nowhere near the kind of readiness and preparedness campaigns of the cold-war of course, but it's a step up in alertness that's higher than anything I can remember.
My personal barometer: when do government insiders sell all of their stock? They have already done all of the insider trading. Fleeing the sinking ship the second they hear of Bad News would be in line with previous behavior.
Most of the rest of them I can wave off as probably incompetence and maybe augmented by LLM-psychosis in decision-makers, and the general level of chaos in IT these days. The NSA internal network being down is pretty suspicious. Maybe the Trump administration rot has just spread that far, but that is making me think there could be actual cyberwarfare going on.
Must be a little bit of both right? Like FAA grounding + Microsoft outage are both notable on their own, but the subsequent things would usually be background noise IMO.
I also think generally that AI is probably both the cause and solution to these problems. We're certainly getting more code churn than otherwise, and in lots of cases that's really good! But you're going to have to deal with inevitable instability too.
With the trilateral summit scheduled for Thursday it's possible some showing off is going on between the three letter agencies in the three countries... after all they can be first movers on vulns that are likely to be patched soon
100%. Everything is rage-bait nowadays. And AI is the newest toy they can use. Every time we've had something go down in the past ~6months, "AI did it" was all over the comments.
Is it possible that the false of productivity is leading to worse products and massive loss in valuable headcount? Instead of AIs directly writing bad buggy code? Or is "everything rage-bait, and AI is all sunshine"
These kinds of things happened before AI. It's not clear whether the incidence has increased since AI, or maybe AI is leading to faster development and therefore higher rates of incidents may be due to shipping more code, faster, while perhaps maintaining a similar or better level of incident to code/progress ratio. Or, there could just be more attention as GP suggests. Pointing to one example of a company that's using AI and has has a major incident doesn't really tell us much, especially when most companies who haven't had major incidents are probably using AI in similar amounts.
It would be interesting if someone could do a proper analysis, though!
I believe it's also possible that a large portion of the world wants this to be true more than it actually is so attribution to things that have nothing to do with AI go directly to AI as a knee-jerk reaction.
If we're also just vibing it (like the majority of code these days) then I vibe that there are a lot of AI-apologists that desperately don't want their magic machine to be cast in a negative light, lest their easy i-win button gets taken away or devalued.
I would assume if anything it is that the open weight models are lagging by 6-8 months and 6 months ago Mythos was announced.
It would hardly be shocking that hostile nations are able to cause some problems with Mythos level tools.
If anything, it is more interesting how we can invade another country and the idea that the country or a collection of hostile countries could hit us back in any way, shape or form is just an impossibility. An insane conspiracy theory.
Maybe it is not that but surely the probablity is not zero.
Aside: I forget the name of the Baader–Meinhof phenomenon when it hasn't been discussed in awhile. When I'm reminded of it I begin seeing it happening frequently for the next few days. I wish I was joking. I assume it's the Baader–Meinhof phenomenon phenomenon.
I've heard the term "frecency bias" a portmanteau of frequency and recency to refer to people focusing intently on a set of outlier events that happen to occur with a heightened frequency very recently.
I'm sure most of us here have been on the receiving end of managers suddenly obsessed with a failure mode because it happened to occur twice in one month.
This situtation reminds me of the East Palestine train derailment in Ohio. A bad train accident caused a big chemical spill, made national news. For months afterward, any issue with any train in the state was reported as news. People thought the trains were breaking down. In reality, train derailments happen pretty often but it's not usually a big deal so nobody hears about it.
I think it lies at the intersection of Occam's and Hanlon's razors. The simplest explanation is incompetence brought on by a lack of skill from cuts, and therefore stupidity not malice.
It's probably both stupidity and malice, especially on the government side. Project 2025 makes sure that malice is there as a starting point and then the current loyalty-driven hiring practices means that stupidity is a key factor in the execution.
Yeah "cyber thing went down" has certainly become a reporting trend. I'm not sure we would have even heard about NSA internal network issues a year ago. May or may not be significant.
I think the opportunity right now where we have powerful AI hacking systems but not enough protections is to establish long-term presence in all networks of interest. Not bring them down.
Completely agree. Things are moving too fast and so are breaking.
For example, we updated a GCP managed base image of Ubuntu that was released with 350 CVEs patched, some were as high as 9.8. In doing so, they (GCP) changed the DEFAULT setting for hardware GRO from off by default to on. That's a massive networking change packed into a security update.
We only caught it AFTER it started causing issues in production. We didn't have the OS pinned because (decently valid reasons) or catch it in non-prod (decent reason + human mistake).
This is on a modern highly automated tested resilient well architected platform.
I can only imagine these other platforms are dramatically more exposed that we are, hence these common large scale outages.
Also, anyone anywhere that horded/designed exploits will be wanting to use their backlog of these "assets" before systems are patched. From criminal networks to nation states.
Outages could be grey-hat testers and emergency updates.
A temporary period like we are experiencing is to be expected when moving to a time where a single engineer can perform the equivalent work of a team of excellent engineers working for weeks. Security has always been a game of who has the capabilities, means, and requirement for any given attack surface.
That assumes that people did not already have long-term presence in networks of interest.
More plausible to me is that spy agencies had previously found and used some of these hundreds or thousands of vulnerabilities to hack into those networks just fine. But now other people like cyber criminals are using these vulnerabilities via AI. That is messing up both the attackers who had already gotten in quietly and do not want the attention and the defenders who now have to respond to a noisy intrusion they cannot ignore.
I have also been curious about frequency bias / Baader–Meinhof - but I am also wondering: what are the chances that a state actor is responsible for the last week of issues?
Trump is headed to China for a meeting with Xi and Putin.
We have 3 carriers off the coast of Iran and (despite Trump's pleas) NATO is essentially pushing their boot into Putin's neck w/ continued Ukrainian strikes on Russian oil infra.
I've seen a few takes from L3 Harris / FAA employees calling the official line about the ATC issue bunk and there are 3 other major issues in the last week, now this NSA outage - curious if these incidents could be the BRICS axis flexing their own cybermuscles (independently in multiple separate attacks with carefully placed assets) ahead of their meeting with Trump in China?
An interesting geopolitical Mexican standoff, if true.
I'm not claiming this is true. I'm literally pulling it out of my rear with little context beyond a bunch of random news stories and X posts, it's pure pedestrian speculation / curiosity.
Would love someone steeped in national security / geopolitics (like: professionally as their job, not armchair analysts) to offer their take on this.
>NATO is essentially pushing their boot into Putin's neck w/ continued Ukrainian strikes on Russian oil infra.
I know it's not the thrust of your post, but I see this or similar framing a lot, and I feel compelled to point out that Putin can end the strikes on his oil infrastructure at any time by ending the war he started by invading another country.
I suppose rather than straight up 'cyberattack', that some kind of sabotage is also possible... but I don't see how that could ever really be proven without massive state-level investigations... at which point people will also question whether they are themselves involved somehow.
Following the series of such events over the past couple of days, I have to say I’m a bit surprised by how quickly some in this community are brushing them off as AI-related pain or simply another example of the clickbait-driven media ecosystem.
I understand the skepticism, and I certainly don’t think we should go down a rabbit hole of conspiracy theories. But shouldn’t we also be a little skeptical of dismissing these events so quickly? Is it unreasonable to at least consider the possibility that they’re not entirely coincidental, or to wonder why so many seem to be taking place in the US?
IMO it's unreasonable not to think they are all related in some way, even if that relation is just bad actors seeing what other bad actors have been capable of recently. This all started with the freezers in Europe and seems to be snowballing since.
I have this distinct memory in I think Jan 2020 where I looked at the front page of CNN or NYT or something. Big picture of Trump impeachment, and proceedings. Small picture at the bottom of the page of China dealing with this virus. Worthy news, but not top page news. Nobody was paying attention. Feb, more people started paying attention, but it wasn't top page yet. By March, people were starting to realize that this is a serious crisis that's going to severely impact their lives.
I'm getting Feb 2020 vibes. I don't know what's going on, but this feels like more than just some routine outages going on.
The fantasy/cyberpunk TTRPG Shadowrun depicted a world in upheaval after a global computer crash in the late 2020s. In addition to the pandemics and the AIs.
It's AI. Or more specifically: rushed, careless usage of AI under the pretense that it's already "solved" software and you can haphazardly push junk to the world and expect real systems to run on it.
Traditional software development normally involves at least a few veterans on the team familiar with almost every line of code.
However, AI works like a speedy intern who hasn't yet got to grips with the whole codebase, and sometimes makes assumptions. That's mostly because the codebase is usually bigger than the context window, so it is 'writing blind'.
Allowing our infrastructure to be manufactured elsewhere, maintained by companies who struggle to retain top talent, and overseen by folks out of touch with cutting edge threats, and funded on annual budget cycles by legislators even more out-of-touch ... is a seemingly insurmountable set of obstacles.
... and Doug's perl script, which was never committed to source control. It lives on a single box and silently upholds a number of implicit system invariants. In someones inbox is an email from Doug, sent just before he retired, that the script needs to be copied over to the new box before the hardware migration is complete.
There's some speculation that all of these recent outages might be related to attempts at taking Genesis Mission online - the Trump / Oracle project to have AI tie into government and private services:
AI is tuned to speak confidently and ship independently; it is rare that it will suggest that one go back to a broad survey of stakeholders to ensure no Chesterton's Fence is being torn down (https://fs.blog/chestertons-fence/). "I've read the documentation and logs, and nothing suggests this system needs these legacy capabilities/components/capacity" is something we hear all day, and it is a siren's song if there ever was one.
Couple that with a general (and valid!) consensus that modernization of legacy systems must be accelerated for a variety of reasons - to mitigate this new and continuing "Black October" of hyper-agile agent-driven cyberattacks (whether the network is public or private); to have more flexible infrastructure for new features to stay ahead; to allow human teams to show their efficacy in a world where technology jobs are increasingly at risk - and you have a perfect storm where change management is no longer what it once was.
The FAA's recent outage for example, at its core, seems like people trying to rush needed upgrades that were slow for a specific reason (e.g. reducing the testing window for 90 to 30-days).
The Microsoft outage seems like three different things linked together (and poor "full failure" recovery prep, lengthening the outage).
The bank "outage" may not have even existed, 2/3 of the banks claimed they didn't have an outage at all, and the other said it was a limited technical issue. The articles were based on DownDetector spikes.
I can't put my finger on it, but something sinister feels afoot.
https://beready.campaign.gov.uk/
It's nowhere near the kind of readiness and preparedness campaigns of the cold-war of course, but it's a step up in alertness that's higher than anything I can remember.
I also think generally that AI is probably both the cause and solution to these problems. We're certainly getting more code churn than otherwise, and in lots of cases that's really good! But you're going to have to deal with inevitable instability too.
https://www.bleepingcomputer.com/news/security/optimizely-su...
Is it possible that the false of productivity is leading to worse products and massive loss in valuable headcount? Instead of AIs directly writing bad buggy code? Or is "everything rage-bait, and AI is all sunshine"
It would be interesting if someone could do a proper analysis, though!
It would hardly be shocking that hostile nations are able to cause some problems with Mythos level tools.
If anything, it is more interesting how we can invade another country and the idea that the country or a collection of hostile countries could hit us back in any way, shape or form is just an impossibility. An insane conspiracy theory.
Maybe it is not that but surely the probablity is not zero.
I'm sure most of us here have been on the receiving end of managers suddenly obsessed with a failure mode because it happened to occur twice in one month.
The combination is a cycle
1. AI finds bugs
2. Engineers us AI to try and fix
3. AI testing approves, minimal engineering review
4. Engineers push release rapidly, it's critical after all
5. Bugs introduced
6. Cycle starts again
Coincidence or is there some common explanation that I missed?
I think the opportunity right now where we have powerful AI hacking systems but not enough protections is to establish long-term presence in all networks of interest. Not bring them down.
For example, we updated a GCP managed base image of Ubuntu that was released with 350 CVEs patched, some were as high as 9.8. In doing so, they (GCP) changed the DEFAULT setting for hardware GRO from off by default to on. That's a massive networking change packed into a security update.
We only caught it AFTER it started causing issues in production. We didn't have the OS pinned because (decently valid reasons) or catch it in non-prod (decent reason + human mistake).
This is on a modern highly automated tested resilient well architected platform.
I can only imagine these other platforms are dramatically more exposed that we are, hence these common large scale outages.
Outages could be grey-hat testers and emergency updates.
A temporary period like we are experiencing is to be expected when moving to a time where a single engineer can perform the equivalent work of a team of excellent engineers working for weeks. Security has always been a game of who has the capabilities, means, and requirement for any given attack surface.
More plausible to me is that spy agencies had previously found and used some of these hundreds or thousands of vulnerabilities to hack into those networks just fine. But now other people like cyber criminals are using these vulnerabilities via AI. That is messing up both the attackers who had already gotten in quietly and do not want the attention and the defenders who now have to respond to a noisy intrusion they cannot ignore.
Trump is headed to China for a meeting with Xi and Putin.
We have 3 carriers off the coast of Iran and (despite Trump's pleas) NATO is essentially pushing their boot into Putin's neck w/ continued Ukrainian strikes on Russian oil infra.
I've seen a few takes from L3 Harris / FAA employees calling the official line about the ATC issue bunk and there are 3 other major issues in the last week, now this NSA outage - curious if these incidents could be the BRICS axis flexing their own cybermuscles (independently in multiple separate attacks with carefully placed assets) ahead of their meeting with Trump in China?
An interesting geopolitical Mexican standoff, if true.
I'm not claiming this is true. I'm literally pulling it out of my rear with little context beyond a bunch of random news stories and X posts, it's pure pedestrian speculation / curiosity.
Would love someone steeped in national security / geopolitics (like: professionally as their job, not armchair analysts) to offer their take on this.
Is this a dumb, naive theory?
I know it's not the thrust of your post, but I see this or similar framing a lot, and I feel compelled to point out that Putin can end the strikes on his oil infrastructure at any time by ending the war he started by invading another country.
Historically he's chosen to go about things differently, which I guess is part of what is tickling the pattern matching part of my brain here.
I'm not sure about naive or dumb but you might review your POV regarding this.
I understand the skepticism, and I certainly don’t think we should go down a rabbit hole of conspiracy theories. But shouldn’t we also be a little skeptical of dismissing these events so quickly? Is it unreasonable to at least consider the possibility that they’re not entirely coincidental, or to wonder why so many seem to be taking place in the US?
I'm getting Feb 2020 vibes. I don't know what's going on, but this feels like more than just some routine outages going on.
So far this fall, none of my systems have gone haywire (yet!), so I forecast that we'll be OK.
However, AI works like a speedy intern who hasn't yet got to grips with the whole codebase, and sometimes makes assumptions. That's mostly because the codebase is usually bigger than the context window, so it is 'writing blind'.
Allowing our infrastructure to be manufactured elsewhere, maintained by companies who struggle to retain top talent, and overseen by folks out of touch with cutting edge threats, and funded on annual budget cycles by legislators even more out-of-touch ... is a seemingly insurmountable set of obstacles.
https://knowyourmeme.com/sensitive/memes/glowie-glowposting
Not unlike Bloomberg terminals…dos era UI but does the job well
Did you try sending it via outlook.com?
https://en.wikipedia.org/wiki/Genesis_Mission
https://substack.com/@thedreydossier/note/c-389611426