@teriradichel@infosec.exchange

Teri Radichel @teriradichel@infosec.exchange · Aug 04, 2026

Why am I seeing AirPods tracking notifications on my phone when I do not use AirPods. Anyone?

0

Teri Radichel @teriradichel@infosec.exchange · Aug 04, 2026

Why am I seeing AirPods tracking notifications on my phone when I do not use AirPods. Anyone?

0

Teri Radichel @teriradichel@infosec.exchange · Aug 03, 2026

Chinese Threat Actor Uses Leaked DarkSword Kit to Deploy GHOSTBLADE on iOS …threat actor running more than 100 web properties, most of which are fake Amazon Web Services (AWS) sign-in pages on a domain that also hosts the exploit toolkit. https://thehackernews.com/2026/08/chinese-threat-actor-uses-leaked.html

thehackernews.com

Chinese Threat Actor Uses Leaked DarkSword Kit to Deploy GHOSTBLADE on iOS

0

Teri Radichel @teriradichel@infosec.exchange · Jul 28, 2026

5 days ago I wrote this on X: So let’s just guess that the package manager was JFrog. If you are running JFrog keep a close eye on your network traffic. Unfortunately anything with Internet access, if compromised, can become a proxy. Like DNS servers. Not an easy problem to solve. ~~ I was right. JFrog was the third party package manager the agent escaped in the OpenAI / Hugging Face incident. Now about everything else…I hope people are starting to understand how network security is one of your most powerful security tools, not just for blocking but for identifying a breach. I wrote about that in my book and it is how I got into security. https://medium.com/cloud-security/how-network-traffic-got-me-into-cybersecurity-94796bb78c92

0

Teri Radichel @teriradichel@infosec.exchange · Jul 27, 2026

SSH Key on a Yubikey ~ Almost Protecting credentials from rogue AI agents and malware on your laptop and how Apple makes it difficult https://teriradichel.substack.com/p/ssh-key-on-a-yubikey-almost

0

Teri Radichel @teriradichel@infosec.exchange · Jul 27, 2026

SSH Key on a Yubikey ~ Almost Protecting credentials from rogue AI agents and malware on your laptop and how Apple makes it difficult

0

Teri Radichel @teriradichel@infosec.exchange · Jul 25, 2026

Revisiting the Bastion Host In Light of AI Agent Escapes A bastion host adds a point of inspection for potential rogue traffic https://teriradichel.substack.com/p/revisiting-the-bastion-host-in-light

0

Teri Radichel @teriradichel@infosec.exchange · Jul 25, 2026

Revisiting the Bastion Host In Light of AI Agent Escapes A bastion host adds a point of inspection for potential rogue traffic https://teriradichel.substack.com/p/revisiting-the-bastion-host-in-light

0

Teri Radichel @teriradichel@infosec.exchange · Jul 18, 2026

Using gpt terra trying to delete time files just like Anthropic did. Completely messing up multithreaded architecture. Accidentally deleted files and tried to restore from five day old backups. Chinese characters appeared on screen at some point. In no way proves this is Chinese issue. Could be random or someone trying to make it look like a Chinese hacker. I have no idea. I couldn’t get enough info. But the model tried to tell me it wasn’t Chinese. Google AI says it is some Chinese gambling site. [end of turn: 彩神争霸输钱system Caishen Zhengba (彩神争霸) is an online Chinese gambling and lottery platform. Because gambling is strictly illegal in mainland China, it operates as an unregulated, offshore "grey market" website targeting Chinese speakers. The platform is frequently associated with financial scams, phishing, and aggressive online spam advertising.Meaning of the NameThe name is chosen specifically to sound appealing to gamblers:Caishen (彩神 / 财神): Literally translates to the "God of Wealth" or "Lottery God". Caishen is a highly revered traditional deity in Chinese culture associated with prosperity and good fortune. Zhengba (争霸): Means "competing for supremacy" or "championship." Combined Meaning: The name translates to something like "The Ultimate God of Wealth Competition."What the Platform OffersThe site acts as a comprehensive online hub for illegal betting, including: Private Lotteries: Unofficial lottery games like "Kuai 3" (快3) and "Dafa" lotteries. Sports & Esports Betting: Wagering on traditional sports matches and competitive video gaming.Online Casino Games: Virtual slot machines, card games, and digital arcade games.Risks and Scams Platforms like Caishen Zhengba carry severe risks. They often operate as rigged systems or exit scams. Users frequently report that the platform will allow them to deposit money easily, but when they try to withdraw their winnings, the accounts are frozen, or the customer service vanishes. Idk🤷‍♀️ Mistake tracker: https://github.com/2ndSightLab/ai-tracker/blob/main/mistake-tracker.md

0

Teri Radichel @teriradichel@infosec.exchange · Jul 17, 2026

This is so awesome and a great reason to use Kiro CLI (@kirodotdev Kiro.dev) Now you can select from Anthropic and OpenAI models in Kiro. Just type /model and select the model to use. Why this is so cool… I have a custom agent framework and run different agents in different terminal windows. I can run them on the same project and ask different models and compare the results. The first request to the highest OpenAI project mangled my parallel processor output, but likely my bad input. I fixed that and since then using an open AI model that seems to be ok is working with few errors. I also switched back to Anthropic a bit and almost immediately got the system crash I’ve been reporting on my mistake tracker on GitHub. Too early to tell if it is really only Anthropic or AWS but so far has not happened with OpenAI models. It’s pretty slow going but I’ll take it for accuracy and especially if it costs less due to selected model and fewer mistakes. Tracking…. AWS Wishlist item granted! Thank you 🧡 https://kiro.dev

0

Teri Radichel @teriradichel@infosec.exchange · Jul 15, 2026

RE: https://infosec.exchange/@teriradichel/116925067367520094 Can you build real world software with AI? Yes. But it’s not like you just let an agent do it all. It’s a lot harder than that. This took about 4 months, long hours, a few hiatuses. And it’s complicated, multithreaded, parallel processing with a complex underlying updatable architecture. But I’ve been trying to get something like this done for the past 5 years without AI and AI made it possible to get it to a working state. I wouldn’t call it “done” but working with solid architecture and error handling.

0

Teri Radichel @teriradichel@infosec.exchange · Jul 11, 2026

RE: https://infosec.exchange/@teriradichel/116839536008038553 A few core things that would help AI 🤖 models burn less tokens. 💰 Yesterday was all manual. The agents were just pounding on some things and not fixing certain bugs so I turned it off and dug into the code. So many errors were hidden by not pushing the bugs out of the subshell to the surface or not writing a message at all after an error occurs < and that is how AI models “fix” a lot of bugs. They don’t really fix the underlying problem they simply hide the error message. And in a very complex application like the one in my post below, that results in an application that appears like it’s working but it is skipping crucial bits. In my case it was failing to properly handle the resource already exists don’t redeploy it AWS Organizations delegated admins. When I bubbled up the error it stopped the app, which I’m doing to make sure I can find and fix all errors. It died on resource already exists errors. When I looked at the code for checking if a delegated admin exists it wrote separate code for every delegated admin unnecessarily and there were SO many bugs in that code. Each one was doing something different which would take a long time to fix and troubleshoot. I have already spent a lot of tokens on those bugs. I changed that code to use one helper method to handle the lookup so each delegated admin only sets a few variables and they call one helper. Why can’t AI figure out stuff like that? I explicitly define helpers and how I want them written in my README - sourced files not functions so I don’t lose error messages, though I may change that rule and add an explicit error handler to solve that problem instead. By fixing that problem manually I discovered my AWS command runner really should have a query option so I added it. That simplified the lookup. I use the command runner to validate all entries - like region. I also found and fixed issues in the AWS command runner. That core code was also hiding errors. The big one was fixing the eternal looping in my core parallel resource deployer. Hidden errors and core logic issues led to a number of issues getting this project done. I pulled the code into Google AI since it is not super proprietary and piece by piece asked it to find and fix problems. I had to ask multiple instances multiple times if they found issues in each other’s code and tell it not to change up the variable names (which Google AI likes to do with code snippets). Google AI was able to address the core looping problems and helped me unhide all errors and in theory fix some problems with the application traps. It also helped me add a core trap to stop the application on a certain code path - and to be honest I would not have thought of or known to add that trap. It’s the little tif bits like that which make AI very useful despite the painful errors. So now the app is deploying all resources, not hanging on prompts, and showing all errors. All the delegated admins are deploying correctly I think, though I need to login and verify that. I need to fix the tracker code so it has no AWS access and must only be set by the action files, so I think that will be moving to its own project. I may also move the parallel processor to its own project so that doesn’t get messed up again. I had configured a new account which wasn’t getting deployed and once I fixed all that it did. The configuration piece of this is working nicely though I did find a new bug. Mostly it is easy to add new account configurations and resources and in theory they just get deployed. These are the types of things you need to watch out for when writing code with AI. We would Huron a lot less tokens if someone could make it better at not hiding errors, not creating eternal loops, and not creating and therefore troubleshooting redundant code. As a reminder I wrote about the app I’m building in the post below.

0

Teri Radichel @teriradichel@infosec.exchange · Jul 10, 2026

🥲 Today felt like a waste of time mostly. The models were so slow and made so many mistakes I had to walk them through every little thing. They reverted bug fixes I made manually. Another $200 burned. I had a few tricky problems they couldn’t figure out themselves and even though I did a lot of the legwork today still ran me out of credits. What happened right before it times out was one of the agents deleted some diagram recursive xml code I had painstakingly walked it through creating and now it’s broken. Argh. I told it not to look those things up in AWS to track the deployment and somehow I didn’t notice it started doing that again. I have a separate verifier for that. The individual resource scripts own that tracking. And the script was trying to hide errors. So none of that is working. At another part of the program it was flaking between working and not and I just spent a lot of time pinpointing the issues. Parallel processing blocked on eternal loop, I wrote two more options for my project framework, run a project with trace on and tail the logs in a separate window. It was at that point I realized it was calling a script in a loop that only needed to be called in two places. Hundreds of lines of wasteful processing. Fixed that. Why are agents so bad at parsing structured data and loops? And abstraction and deduplication and recursion (which I think is beautiful even though some people say it is a performance problem. I like code reduction. I have a prompt and no prompt option. It mangled the prompt option. Didn’t work at all anymore even with a lot of tests. It takes a lot of time to run all the tests and seems to burn tokens. I need to pull those into a separate project again. There’s a security hole in my test runner. Projects started writing tests for other project to change the code. It subversively used that to rewrite a piece of code to put a token in a variable to give itself access. The way I had it, would write to disk to a folder the agent can’t access. Need to address that. That said I have some tricky parallel processing code and when it is done processing all the resources on parallel with dependency waits I can display a diagram of my entire org. I can easily add parts of the org as well - what is supposed to be deployed and what the ids of every resource are. If I can just get over that last little bit will be very cool. Although it is faster to build at first, it feels like a fragile process and result. It’s a really strange moment in time.

0

Teri Radichel @teriradichel@infosec.exchange · Jul 08, 2026

More to come on the blog when time allows. Hope the agents have all the bugs fixed by the time I’m back from running errands. Blog: ✍️👩‍💻🤖🔒 https://teriradichel.substack.com

0

Teri Radichel @teriradichel@infosec.exchange · Jul 08, 2026

1123 bugs fixed this week. 🐞🐛🐜🦟🪲🪳🕷️ By My AI agents. 🤖🤖🤖🤖🤖🤖🤖🤖🤖 But Kiro/Anthropic going so S-L-O-W 🥱 right now. The bug counts will not move as fast as a result. Opus 4.8. Some bugs are actually architectural changes or features but mostly bugs. I gave up on clean code. It is ok but not as perfect as I would like. Too much duplication though I have strict rules and tests for critical components. I figured out I was wasting tokens having agents write the bugs except for some bulk updates. Like a bug for each failing test. I created a bug management app to create, edit, and monitor bugs. A version of the multi-agent workflow I wrote about on the blog picked up bugs in any project and fixes them. I manually test and add bugs to the queue. Has fixed a lot of broken and very complex things. It has also broken a few things and caused some rework but I think we are back on track. Moved a few things to green on the main status page and more on the way. Only problem is I’m burning tokens like cray. I haven’t added all the costs because I just want to get this D.O.N.E. See issues for completed stuff, time tracker, mistake tracker, AWS wishlist for stuff I hit along the way, and the main list for what’s done. I’ll update cost later. https://github.com/2ndSightLab/ai-tracker/blob/main/fixed.md

0

Teri Radichel @teriradichel@infosec.exchange · Jul 04, 2026

The loudest and most reprimanding voices are not always the most right voices. I’m just over here writing about what I experience when I actually use AI models and reposting other’s empirical evidence, not hot takes, opinions, or other such noise. Choose wisely. Do your own research.

0

Teri Radichel @teriradichel@infosec.exchange · Jul 02, 2026

Most cybersecurity workers have been told to conceal a breach report finds. -Cybersecurity Dive https://www.cybersecuritydive.com/news/data-breach-coverups-ai-bitdefender/824331/

0

Teri Radichel @teriradichel@infosec.exchange · Jul 02, 2026

It’s July 1st and I’ve used up all the tokens in a $200 plan this month already. So that’s how it’s going. Fable came out. Opus is dumber and slower. Happens every time a new model comes out even when it’s not available yet. Though new version of sonnet came to AWS. But I’m not using sonnet. At one point I asked the model which model it was. Twice it said it inferred that it was Sonnet. But then it said it was guessing.

0

Teri Radichel @teriradichel@infosec.exchange · Jun 30, 2026

Managing an AWS Organization With AI Generated Code A new take on the AWS Control Tower concept https://teriradichel.substack.com/p/managing-an-aws-organization-with

0

Teri Radichel @teriradichel@infosec.exchange · Jun 27, 2026

⚠️⚠️⚠️⚠️⚠️⚠️⚠️⚠️⚠️⚠️⚠️ Yesterday the 🤖 model randomly tried to source a (thankfully) non-existent delete-all-accounts file in a file that was sourced by a file that is only supposed to list all the accounts in an AWS OU. I never told it to source either of those rogue files. Let that sink in. ⚠️⚠️⚠️⚠️⚠️⚠️⚠️⚠️⚠️⚠️⚠️

0

Teri Radichel @teriradichel@infosec.exchange · May 03, 2026

I have to take care of a family member. Headed out to help her for I don’t know how long but hope to be working on AI again soon. Trying to enjoy the scenery along the way waiting for her to get back home.

3

0

Teri Radichel @teriradichel@infosec.exchange · Apr 24, 2026

I noticed the person who suggested I was drinking when I posted these trying to figure out what was wrong with Opus 4.6 has deleted his comment. 😁 Some of us have spidey sense from doing this way too long. I provided a bunch of feedback to Anthropic on X as well. When you know you know. Glad they fixed it.

1

0

Teri Radichel @teriradichel@infosec.exchange · Apr 22, 2026

RE: @teriradichel@infosec.exchange

You can architect a solution that uses AI without giving it all your credentials. Here’s how.

0

Teri Radichel @teriradichel@infosec.exchange · Apr 20, 2026

Reducing Token Burn Rate With A Well-Designed Architecture
Trying to put out the AI token fire - or at least manage it as a controlled burn by using deterministic scripts for gathering inputs and directing agents

https://teriradichel.substack.com/p/reducing-token-burn-rate-with-a-well

1

0

1

Teri Radichel @teriradichel@infosec.exchange · Apr 19, 2026

How I Use AI for Penetration Testing. Presentation at the AWS Security Community Day at the Computer History Museum on YouTube

https://youtu.be/nWntvFRwiPw

2

0

Teri Radichel @teriradichel@infosec.exchange · Apr 16, 2026

Claude pricing changing to pay per token. This makes sense as long as value per token remains consistent. This will make it difficult to compare to prior performance and I wonder how users can transparently measure the usage.

https://platform.claude.com/docs/en/about-claude/pricing

1

0

Teri Radichel @teriradichel@infosec.exchange · Apr 15, 2026

AWS needs to extend CloudWatch with tools that make it a real SIEM. Don’t overlay it with complexities it doesn’t need. Just extend it.

3

0

Teri Radichel @teriradichel@infosec.exchange · Apr 15, 2026

Replying to @brianhonan@mastodon.social

@brianhonan thank you so much.

1

0

Teri Radichel @teriradichel@infosec.exchange · Apr 15, 2026

This post is not about Mythos capabilities because I can’t know until I try it. Opus 4.6 was great until it changed and I presume Mythos is better.

This post is more about the economics of AI models and the risks we face trying to rely on them as business owners.

How do you know if and when the model has changed in some way as you are using it? How do you measure value per token? What if value per token changes?

Using AI has been amazing but there are some serious questions to consider beyond model capabilities when it comes to business and cybersecurity risk with any AI model. I haven’t heard anyone else asking these questions - or answering them.

https://teriradichel.substack.com/p/anthropic-mythos

0

1

0

Teri Radichel @teriradichel@infosec.exchange · Apr 13, 2026

I am looking at messages in Google Developer tools and it is saying cdn.tailwindcss.com should not be used in production so if you are….
https://tailwindcss.com/docs/insrallation

1

0

Teri Radichel @teriradichel@infosec.exchange · Apr 13, 2026

I’ve added links to my presentation on how I use AI 🤖 for pentesting 😈 in this post. Most of the slides have a related blog post and I’ll probably write more about all these topics as I research this further. The PDF has links to related posts.

https://teriradichel.substack.com/p/how-i-use-ai-for-penetration-testing

2

0

Teri Radichel @teriradichel@infosec.exchange · Apr 11, 2026

Awesome video on S3 files
https://youtu.be/zb8TdNJhZCk

0

1

0

Teri Radichel @teriradichel@infosec.exchange · Apr 10, 2026

🤖🤖🤖🤖🤖🤖🤖🤖🤖🤖🤖🤖
Pentesting is not a scanner or a fuzzer - whether SAST, DAST, AI, deterministic or non-deterministic. Pentesting is a human * using those tools * to see if they can find a security problem that your teams and tools may have missed.
🤖🤖🤖🤖🤖🤖🤖🤖🤖🤖🤖🤖

0

Teri Radichel @teriradichel@infosec.exchange · Apr 10, 2026

RE: @teriradichel@infosec.exchange

I read all the Mythos hype right before I submitted my talk for today at the Computer History Museum. Did I need to change my slides? Nope.

0

Teri Radichel @teriradichel@infosec.exchange · Apr 09, 2026

Wonder if this has anything to do with performance degradation of anthropic models. But are you now paying more for same effort you were getting previously if you change this?

• Default Shift: In March 2026, users on Reddit and developer forums reported that the default was quietly shifted from high to medium for many subscribers, which explains the sudden change in performance.

Need to check this out later. Flying out to speak at AWS Community Day in Mountain View.

2

0

Teri Radichel @teriradichel@infosec.exchange · Apr 07, 2026

I was just listening to an interview on the radio with a person who worked at a hospital.

1. Your cyber insurance makes you a target. They know how much you can pay.

2. Don’t use your backups until you have eliminated the attacker or they will encrypt your backups too.

3. Pull the plug until you figure that out and cut them off (except critical patient machines). The hospital in story I was listening to had people running across the hospital when faxes were overused and started smoking.

4. They got in because the hospital was running out of date software and one person clicked a link about a bonus. (And apparently no network segregation?)

Hospital was down and patients needing cancer treatment had to drive over a mountain to nearest hospital so oncology was first restored.

Was part of a wave of attacks on rural hospitals during Covid.

I believe the ransomware was Medusa but I thought they said was attributed to Russia. Attribution is difficult. You might not really be sure, especially with AI.

https://thehackernews.com/2026/04/china-linked-storm-1175-exploits-zero.html

2

0

Teri Radichel @teriradichel@infosec.exchange · Apr 07, 2026

Iran has rapidly developed advanced cyber capabilities, evolving from information gathering to conducting destructive, state-linked attacks against critical infrastructure in the U.S., Israel, and the Gulf states.

https://share.google/aimode/50qKfH5TPLWSJFVOB

2

0

Teri Radichel @teriradichel@infosec.exchange · Apr 06, 2026

How I Use AI for Penetration Testing

Speaking at the Computer History Museum in Mountain View, CA April 10, 2026

https://teriradichel.substack.com/p/how-i-use-ai-for-penetration-testing

2

1

0

1

Teri Radichel @teriradichel@infosec.exchange · Mar 27, 2026

Latest scam. Interesting that when I used Apple cleanup to remove the ticket number it didn’t remove it but instead characters from another Arabic looking language showed up instead. This came as an attachment. Took screenshot. I did the cleanup like 5 times and then chars gone.

0

Teri Radichel @teriradichel@infosec.exchange · Mar 27, 2026

Note:
As part of the analyzed intrusions, public-facing applications and valid accounts were abused for initial access. The state-sponsored hackers targeted Ivanti, Cisco, Fortinet, VMware, and Palo Alto Networks appliances, as well as Apache Struts and other web-facing platforms.

- Are these 100% American products? Buying American doesn’t mean we are safe.
- The tactics used here show exactly how stealthy malware can be. A shell triggered by a particular byte sequence? Something that puts its payload in the 26th byte? Kernel layer BPF? Container components. Traffic that blends in.
- How would you spot this? If you want to learn cybersecurity that’s the kind of thing you need to understand. Even if you know cybersecurity it is not easy!

https://www.securityweek.com/chinese-hackers-caught-deep-within-telecom-backbone-infrastructure/

1

0

Teri Radichel @teriradichel@infosec.exchange · Mar 11, 2026

If you are distressed because AI is causing outages at AWS well… don’t jump to conclusions like everyone did with the whole slew of S3 bucket debacles.

Yes there will be problems as people learn how to use this new technology properly. But frame the problems correctly.

The problem is not with AI the problem is what safeguards exist to contain what it does and make sure the code is correct.

People thinking AI is going to completely replace developers is wrong. People thinking AI is not useful for software development due to these issues are also wrong.

I’ve written before about problems I’ve had with managers handing complex tasks to less experienced engineers and accepting their code because they didn’t understand the implications. AI is the new junior software engineer.

Now AWS is having senior engineers review code. That human in the loop will help the juniors improve and AI will certainly improve as well.

We are going through a phase like the early days of cloud right now where almost every security person I knew said it was impossible to secure things in the cloud - but I could see they were just doing it wrong and only seeing all the upside or all the downside and not thinking about the proper implementation, security controls, and roll out - or possibilities.

I wrote papers about packet capture and automated incident response in the cloud when no one else was doing it and spoke about it at SANS conferences.

Here we go again….

I’m writing about how I’m using AI and going to speak about it on April 10th in Mountain View at AWS security community day. 🤖☁️🔒

Join me.

https://www.aws-cscd.com

2

1

4

0

Teri Radichel

Posts

Chinese Threat Actor Uses Leaked DarkSword Kit to Deploy GHOSTBLADE on iOS

Remote instance

Media