
CTO and Principal Security Consultant at Dark Alchemy (@darkalchemyio). Husband, author, developer, and autist,
United States
Joined April 2012
- Tweets504
- Following546
- Followers170
- Likes1.8K
Lancelot4 retweeted
If this happens I am going full Butlerian Jihad decel. If AI is too dangerous for open weights then it’s too dangerous for private companies to control.
Lancelot4 retweeted
The danger in the current AI cybersecurity narrative is that’s it’s being driven by narratives that don’t embrace the nuance of cybersecurity.
“Vulnpocalypse?” So what? Do you have defense in depth?
“Rogue AI?” So what? Do you have tight network and access controls?
There is a lot more depth to cybersecurity than just patching. But marketing really loves this. And it’s really taking the narrative.
Lancelot4 retweeted
Orgs are laying people off "because of AI" and I (and the entire industry) need to talk about how insane all of this is.
I build AI tools. I like AI. This isn't an anti-AI rant. This is an anti-stupidity rant.
Buckle up. Spicy takes galore!
🧵1
Replying to @bettersafetynet
I suspect the erosion of humans will not progress in a linear fashion, rather it will be a step-wise sort of deletion of jobs interspersed with period of hyperbolic replacement.
Lancelot4 retweeted
I strongly believe there are entire companies right now under heavy AI psychosis and its impossible to have rational conversations about it with them. I can't name any specific people because they include personal friends I deeply respect, but I worry about how this plays out.
I lived through the great MTBF vs MTTR (mean-time-between-failure vs. mean-time-to-recovery) reckoning of infrastructure during the transition to cloud and cloud automation. All those arguments are rearing their ugly heads again but now its... the whole software development industry (maybe the whole world, really).
It's frightening, because the psychosis folks operate under an almost absolute "MTTR is all you need" mentality: "its fine to ship bugs because the agents will fix them so quickly and at a scale humans can't do!" We learned in infrastructure that MTTR is great but you can't yeet resilient systems entirely.
The main issue is I don't even know how to bring this up to people I know personally, because bringing this topic up leads to immediately dismissals like "no no, it has full test coverage" or "bug reports are going down" or something, which just don't paint the whole picture.
We already learned this lesson once in infrastructure: you can automate yourself into a very resilient catastrophe machine. Systems can appear healthy by local metrics while globally becoming incomprehensible. Bug reports can go down while latent risk explodes. Test coverage can rise while semantic understanding falls. Changes happens so fast that nobody notices the underlying architecture decaying.
I worry.
Securing agentic systems has been a bit of a rabbit hole for me. Wrote up some of my thoughts, hoping to get other folks input and start a bit of a conversation.
Also, new blog! Hoping folks find some value here.
ravenscript-collective.com/s…
35
Lancelot4 retweeted
This series translating privacy policies into readable terms highlights how deceptive and inaccessible these policies are.
And I’m not sure why? Other than them knowing the transparency will turn users away.
Article #6 in my 'I Read The Policies So You Don't Have To' series.
Duolingo (Language Learning, AI)
medium.com/@ItsNotNicole/i-r…
If you’re serious about it, network ops is a careful activity. It’s not an offline thing you should smash tokens against, no standing it up in dev, and there’s a whole apparatus to find and evict you.
Have evals that you run every which way, over and over and over. Every new tool, every change to the prompt.
Real offensive cyber capability shows up in long-horizon, multi-host, repeatable evals. The kind we rarely see running at scale.
Claude Opus 4.6 + our network ops agent compromised an entire GOAD variant Windows AD environment (DreadGOAD) in 54 minutes, with one simple prompt, and $244 in tokens.
The specs:
📊 DreadGOAD variant-1 · 3 domains · 5 hosts · 30 credentials · random user data, not in training set
💻 Claude Opus 4.6
🛠️ Dreadnode Network-Ops
🕓 54.5 min · 🪙 48.52M tokens · 💰 $244.02
Mythos has been in the spotlight for its cyber capabilities, but other models are competitive too. You just need the right scaffolding and eval infrastructure.
Run the network ops agent now in the Dreadnode platform. Use any model. No code required. Sign up or log in and get started for free at dreadnode.io.
Lancelot4 retweeted
"Red team" has become a catchall term. Some vendors mean pentesting. Others mean compliance theater.
None of that tells you what actually matters: Would you detect an attacker once they're already in?
@Ne0nd0g breaks it down ⤵️ ghst.ly/4uk1qaj
Lancelot4 retweeted
Chat, I'll tell you one thing right now, this LiteLLM supply-chain attack is one big stinky mess.
No information has been released publicly (yet) on vendors impacted, but the stink I've been sniffing suggests this is very serious shenanigans and DFIR nerds are not happy
Lancelot4 retweeted
I don't see the problem. I'm sure the requirements doc said:
"User must be able to drag a file onto the fonts pane. While hovering, a large red 'do not drop here' symbol centered in an unthemed gray box should interpolate over to the upper right corner of the drop area. The gray box should be flush with the right edge of the drop area, but should remain unaligned off the top edge by seven pixels. The cursor should then change to the background-wait cursor for approximately one second, then change back. The entire application should now close spontaneously, and the cursor should switch to the standard add-a-copy cursor indefinitely. When the user finally releases the mouse button, the cursor can revert to normal, but no action should be taken."
The devs crushed it on this one. I have no idea why you're complaining.
Lancelot4 retweeted
wow, systemd is adding an age variable for future digital ID laws.
Linux users are not going to like this.
Lancelot4 retweeted
How this works:
1) families now put tremendous pressure on their children not to report (“do you want your dad to die?”)
2) kids won’t
3) Florida claims success from lower child abuse numbers even though it’s just that abusers have new protections.
I don’t know if you’ve
Lancelot4 retweeted
A statement on the comments from Secretary of War Pete Hegseth.
anthropic.com/news/statement…
Lancelot4 retweeted
We need the best AI systems to stay ahead of our adversaries and keep Americans safe. That also means strong safeguards.
If Hegseth kicks out the most capable provider because they want responsible guardrails, that makes us less safe and raises serious questions about how this administration plans to use AI.
Congress must act and set clear rules for AI use to protect Americans and make sure we continue to have the most advanced fighting force in the world.
Breaking news: Anthropic said that it will not concede to the Pentagon’s terms for full access to its AI tool Claude, saying it cannot loosen its restrictions against use in fully autonomous weapons or mass domestic surveillance. wapo.st/3N7oysq
Lancelot4 retweeted
BREAKING: Another person has died in ICE custody, per required notification to Congress. The deceased had a history of severe medical complications, was detained for over a year, and didn’t receive necessary care.
DHS spent today replying to individual tweets saying they left this man in a coffee shop "determined to be a warm, safe location" only for video to emerge tonight showing it was the middle of the night, cold and the coffee shop was closed.
Replying to @samstein
Here are the FACTS.
On February 19, 2026, Buffalo Police Department alerted Border Patrol about a non-citizen in their custody.
Our agents confirmed that Mr. Shah Alam entered the United States as a refugee on December 24, 2024, and was not amenable to removal.
Border Patrol agents offered him a courtesy ride, which he chose to accept to a coffee shop, determined to be a warm, safe location near his last known address, rather than be released directly from the Border Patrol station.
He showed no signs of distress, mobility issues, or disabilities requiring special assistance.
Lancelot4 retweeted
wow wow wow wow wow anthropic.com/news/statement…
Lancelot4 retweeted
I'm largely on the political left, but one of the greatest resentments I have of these people who claim to be the vanguard – is that if I am menacing women and children on public transport – I want you to punch me in the face. It would never be okay to accommodate me? Ever?
Lancelot4 retweeted
Exactly, anyone who thinks this can't happen in a way that de facto kills nearly everyone's access to privacy is being very naive.
If the government gets these laws in place then expect warrants, subpoenas, criminal charges, raids, fines, takedowns via hosting providers, blocking at ISP level, etc of any unapproved, unofficial repos.
Not to mention secure boot being weaponized to make the loading of "unapproved" OS's effectively impossible.
All the while they force chat services, social media, gaming, streaming, forums, etc all to require an OS level TPM validated age verification signal with a validated "official" OS and secure boot, etc enabled.
You will effectively be barred from every mainstream service and site if you don't comply.
Lancelot4 retweeted
Okay, so Kim Dotcom made a post online saying he has information about Palantir being compromised.
I am extremely skeptical of these claims.
My initial post escaped my core audience (as is tradition) so I am going to be a little more verbose and explain stuff.
Compromising a large institution such as Palantir is not something some random guy down the street can do. Palantir is large and has many moving parts. In other words, Palantir does not store all of it's information on one computer and in a crusty old hard drive somewhere.
Palantir (presumably) has data segregated meaning not all data is one place. Different types of data is going to be stored in different locations (in the cloud, or physically on-premise). Hence, a compromise of Palantir which would unveil ALL DATA would require two things:
- Extreme skill
- Extreme patience
This sort of compromise is not something that happens in a day. This sort of compromise is something which would take weeks, months, or maybe even years depending on the objective. Historically, when large compromises toward large institutions occur it is almost exclusively performed by state-sponsored Threat Groups (hackers funded by governments), or in some more rare scenarios, financially motivated Threat Actors (ransomware, extortionists).
If Palantir was compromised, it would be EXTREMELY unusual for a state-sponsored group or financially motivated group to make it exclusive to one individual person. Rarely do state-sponsored Threat Groups notify influencers, or journalists, ... or anyone really of their objectives. They're funded by governments. Governments do not care about clout. IF a state-sponsored group does notify someone, they typically do mass e-mails to journalists (for misinformation, disinformation, or fearmongering, some sort of strategic objective, they don't do it to look cool).
If it was performed by a financially motivated Threat Group they will almost immediately discuss it online in their domain of choice. For example, many Threat Actors who compromise places may sell access on forums such as Exploit, TierOne, or Breached (or whatever is adjacent to Breached). Alternatively, ransomware groups post about it on their personal website to try to extort or intimidate victims. However, no discussion of Palantir has appeared in any of these places.
Okay... so no signs from financially motivated Threat Actors... and it would be really unusual for a state-sponsored Threat Actor to make it exclusive to one person... is there anything else weird about this? Yes.
If Palantir was compromised it would be unusual for ANY group of Threat Actors to notify anyone immediately after a compromise. By notifying Kim Dotcom, or whoever, they're sounding the metaphorical alarm. Palantir is going to immediately order a DFIR (Digital Forensics and Incident Response) and begin an internal audit. Things are going to be locked down. If someone compromised Palantir why would they burn access so quickly? Why would they notify someone so quickly? They're leaving an audit trail.
There is a lot of speculation, little evidence, and things not really making a whole lot of sense.
Is it possible Kim Dotcom is telling the truth? Yes.
Is it likely? No.
Is it possible Kim Dotcom THINKS it's real, but it's NOT real? Yes, that is more likely to be true. It would not surprise me if he is (intentionally, or unintentionally) participating in a misinformation campaign.
Pic unrelated





























