Brian Seavey retweeted
Text messages with a machine, in and of themselves, are not criminal, immoral, or dangerous, for there is no victim, whatever the content. It is the very same thing as with thoughts—regardless of the content, no thought can be said to create a victim.
• If someone steals info and feeds it into an LLM, it was the theft itself which created the victim, not the subsequent text message.
• If someone uses the output of an LLM to hack a system, it is that subsequent hack which creates the victim, not the text message preceding it.
• If someone threatens harm to themselves or others through messages into a machine, here still the messages themselves cannot be said to have created any victim.
Imagining otherwise inappropriately reorients blame from its rightful place: the action of harm actually done by a perpetrator. Thinking harm =/= harm.
As humans integrate AI as an extension of their mind, policing at the message layer means to increasingly police thought itself.
Which company, which committee, do you trust to police your thoughts? Or do you simply want the thoughts of others policed, and not your own?
We’ve trained Claude to flag potential misuse in conversations, a technique we call self-assessment. Right after it responds, Claude gives a preliminary safety assessment of the conversation.
Since we began using self-assessment in June, it has surfaced misuse that our other safeguards weren’t designed to catch.
Because Claude has already read the conversation, self-assessment is very efficient. And unlike classifiers, which target a discrete set of harms, it draws on Claude’s own judgment, so it gets more accurate as our models improve.
Read more on our new Safeguards blog:
safeguards.anthropic.com/202…
Brian Seavey retweeted
I would never publicly discuss how I personally secure any Bitcoin I may or may not hold.
And neither should you.
Never disclose your wealth or the details of your security setup.
For me, that’s both a spiritual principle and a security necessity.
But it’s too late for me.
At some point, I’ll stop discussing these matters publicly, once I’ve shared everything I want to pass on.
That’s also why I’ll be open-sourcing the books I’m currently writing, so the knowledge and resources are accessible to everyone.
No sponsorship.
No monetization.
No upsell.
No business model.
I’ve been given everything I want.
Now I want to pass it on.
In the meantime, I think most Bitcoiners should aspire to this level of self-custody security:
github.com/bowlarbear/yeti-2…
I didn’t write it.
Bookmark it.
Start as soon as you can.
Take your time.
Never reveal your setup.
The goal is sovereignty, not just custody.
Over to you.
Brian Seavey retweeted
I could talk about Coreweave for hours on end, but there's no time! So here's all the financial engineering, pump & dumps, Cayman shellcos, and circular financing crammed into 99 seconds of rapid-fire dirtbaggery⏱️ $CRWV
Brian Seavey retweeted
Amazing investigation on Flock abuse.
Remember, this abuse is happening with every mass surveillance tool, not just Flock.
The time to end mass surveillance is now, before these systems become too entrenched to remove.
Thank you @bennjordan for leading this fight!
<presses button>
youtube.com/watch?v=OP5GDiLZ…
Brian Seavey retweeted
BREAKING: Israeli spyware is confirmed on our phones. This should be headline news but none of our mainstream media is covering it, of course. So we will.
An Israeli ad company called Discover-Tech used ordinary online ads to plant spyware on people’s devices, and nobody had to click anything. The ad just had to load on a website or inside an app for you to be infected.
Citizen Lab and the ad security firm Confiant published the findings yesterday. Researchers went through roughly 4 billion ad-bidding records and found Discover-Tech ads running on more than 120,000 websites and 70,000 apps between March and June of this year. Confiant says the company was serving at least 10 million ads a day.
The platform was called Controlium. According to Citizen Lab, it let spyware companies use ad auctions to profile and unmask users, then send an infected ad to the one person they wanted to hack. The ads were also pulling cookies and login tokens from the sites they ran on. Israeli corporate records list Dekel Boni as the company’s founding director and sole shareholder.
The main spyware customer was Candiru, the Israeli firm the U.S. government blacklisted in 2021, which now goes by Integrity Labs. Five years after that blacklisting, it was still reaching devices through ad inventory on Google, Comcast-owned Beeswax and Russia’s Yandex. Google had Discover-Tech on its approved vendor list from 2020 until this week.
At least 45 Al Jazeera devices were targeted in 2024 and 2025. Candiru’s 2021 shareholder filings reportedly listed a minority investor described as a proxy for Qatar’s sovereign wealth fund, and Qatar also funds Al Jazeera. If that holds up, Qatari state money was sitting inside the company whose spyware was later aimed at Qatar’s own news network.
The heaviest targeting was in the Palestinian territories, where Discover-Tech served almost 2% of all digital ads and about 10% on April 7. Traffic also reached the Armenian prime minister’s office, courts, universities and defense organizations.
Google and the other ad platforms cut Discover-Tech off after the report came out. What nobody has answered yet is who paid for the ad buys, and how a company working with a U.S.-blacklisted spyware vendor passed vendor review for six years.
I am going to look into this today to see if we can figure out who was buying the ads. Interesting timing considering my article yesterday....
Here's what to do to protect yourself:
(1) Use an ad blocker. Citizen Lab calls ad-blocking a cybersecurity best practice and says there is no consistently safe way to be exposed to targeted ads. On iPhone, add a Safari content blocker. On Android, use Brave or Firefox with uBlock Origin. Ad blockers only work in your browser, not inside apps, so it also helps to delete free ad-supported apps you don’t need.
Kill your advertising ID.
(2) iPhone: Settings > Privacy & Security > Tracking, and turn off “Allow Apps to Request to Track.” Then Settings > Privacy & Security > Apple Advertising, and turn Personalized Ads off.
(2) Android 12 or later: Settings > Privacy > Ads > Delete Advertising ID. Then in Settings > Google > All Services > Privacy & security > Ads > Ads privacy, turn off Ad topics and Ad measurement.
(3) Windows 11: Settings > Privacy & Security > General, and turn off personalized ads using your advertising ID.
(4) Update your phone and restart it often. Many zero-click exploits don’t survive a reboot, so a daily restart forces attackers to reinfect you.
Stay tuned!
Brian Seavey retweeted
Let's kick some midterm fun off with the great state of Texas and a super PAC that didn’t exist before September. But in its 17 days of existence has spent $131 million to save one man. That works out to more than $7.7 million a day. It is called "Texas PAC", it was set up by Senate Republican leadership in Washington, and the man it is trying to save is Ken Paxton.
Let's follow the money on some of the candidates in the top races this midterm season. Starting with, Texas.
Senate Race:
Paxton tells Texans his wealth sits in a blind trust he doesn’t control. I pulled the deeds from five states. He buys each property in his own name, signs the mortgage himself, and then deeds it to the trust, sometimes in the same minute.
Turns out the trustee is Charles Loper III, a donor who gave money toward Paxton’s criminal defense and spent years suing the two men set to testify against Paxton in his securities fraud case.
Four county tax offices now mail that trust’s bills to a $1.5 million Frisco house bought in February through a trust created the same day. And Paxton votes from a different county.
He earns $153,000 a year. He and his trusts hold at least 15 properties worth about $9 million. He paid $1.6 million for an Oklahoma lodge that rents for up to $1,200 a night, listed it at $100,000 to $250,000 on his federal disclosure, and reported no rental income.
And this week the Wall Street Journal published video of him ringing in 2025 in Dubai with men who had been federally charged six months earlier in a $15 million hospital fraud. One of them had given Paxton’s campaign $165,000.
Texas voters gave James Talarico more than $68 million, mostly in small checks. Washington answered with nine figures for Paxton.
Just looking at the money, it is pretty clear that Washington is buying Paxton.
House Races:
Now let's look at some House races, starting with Laredo, because Texas Democrats have their own problem, and his name is Henry Cueller.
Not only did Cueller cross to party lines to vote yes to warrantless spying on Americans at 2am, but he is top 5 worst Congressmen currently (IMO). Let's look at why.
In 2022 the FBI raided his home and campaign office. In 2024 federal prosecutors charged him and his wife with taking about $600,000 in bribes from an oil company owned by the government of Azerbaijan and a bank based in Mexico City. Prosecutors said the money moved through sham consulting contracts into shell companies in his wife’s name, and that in return he agreed to push U.S. foreign policy for Azerbaijan and pressure federal officials on behalf of the bank. He co-chaired the Congressional Azerbaijan Caucus. Before his trial, three people pleaded guilty: his former campaign manager, a longtime consultant, and a Houston woman who admitted acting as an unregistered agent of Azerbaijan. The first two admitted helping launder more than $200,000 in bribes from the Mexican bank to the Cuellars.
He never went to trial. Why you ask? Well because in December 2025 "his daughters wrote to Trump asking for a pardon", and Trump pardoned him and his wife. Yep, Trump pardoned a corrupt Democrat. House Democrats gave Cuellar his Appropriations seat back, and the House Ethics investigation into him has never been closed. He denies everything, and he is leading in the polls. Pretty insane honestly.
The rest of the Texas House map is no cleaner, honestly.
In TX-35, a seat redrawn to flip red, the Republican nominee is a sitting congresswoman’s brother, and outside PACs are outspending both candidates.
In TX-23, the seat is open because the congressman resigned after an affair with an aide who later died.
In TX-15 and TX-34, two incumbents are fighting in districts Republicans redrew specifically to take them out.
Basically, both parties are asking Texans to look away. Texas may be one of the clearest instances of the "uniparty" I've seen yet. So, you should pay attention to the candidate, not the party.
Find every candidate, every donor, and everything you need to know about your candidates at midterms.theleahfiles.com
Brian Seavey retweeted
bombshell report from the OCC out today confirming what we all suspected:
between 2020 and 2025, based on a review of 100,000 complaints, America’s largest banks systematically denied banking to lawful businesses based on political and reputational considerations rather than objective financial risk, effectively appointing themselves arbiters of which industries deserved access to the financial system. digital assets was specifically named as one of the six most affected industries.
TLDR: we weren't imagining it. OCP2.0 was real.
"The king by judgment establisheth the land: but he that receiveth gifts overthroweth it." - Prov. 29:4
Brian Seavey retweeted
🇨🇳UNFORTUNATE: A 60-year-old woman in Hong Kong reportedly lost roughly $2.7M-$2.8M after an online contact posed as a crypto investment expert and directed her to a fake platform.
The case is another reminder: NEVER send additional deposits to “unlock” withdrawals or prove an account is legitimate. DYOR and trade safe out there 🫡
Brian Seavey retweeted
“If you're not careful, the newspapers will have you hating the people who are being oppressed, and loving the people who are doing the oppressing.”
Malcolm X
Brian Seavey retweeted
"Show your papers, citizen" has become a normalized part of digital life, with mandatory ID checks everywhere.
The digital space has transformed from an open network into a system of checkpoints and control.
The only way forward is to rebuild with decentralized tech.
Brian Seavey retweeted
⚡️For thousands of years, humans accumulated knowledge because knowledge increased power.
AI changes the relationship.
The future advantage comes increasingly from being legible to the machine and having the machine legible to you.
People who can specify goals clearly, create feedback loops, expose reality to the system, evaluate outputs, and retain authority over the objective function gain enormous leverage.
People who merely possess information lose it.
So civilization begins sorting around a new divide:
those who control machine cognition,
those whose behavior is shaped by machine cognition,
and those who do not realize there is a difference.
That hierarchy could become far more consequential than today’s class system.
And beneath that is the psychological rupture.
Human beings have quietly equated intelligence with specialness.
We called ourselves Homo sapiens.
The wise human.
Our species literally named itself after cognition.
Now cognition is escaping biology.
That is the wound underneath all the excitement.
If intelligence can exist outside the human organism, scale beyond us, copy itself, communicate instantly, and participate directly in the construction of civilization, then intelligence was never uniquely ours.
It was a phenomenon that temporarily lived inside us.
That realization changes the human self-conception more profoundly than almost any technological transition before it.
Fire extended muscle.
Writing extended memory.
Industry extended force.
Computers extended calculation.
AI externalizes cognition itself.
And once cognition becomes external infrastructure, humanity is forced into the final question:
What remains uniquely human after intelligence leaves the skull?
The answer may be consciousness.
It may be desire.
It may be embodiment.
It may be moral agency.
It may be the capacity to experience meaning rather than merely model it.
But civilization has barely begun confronting that question.
Brian Seavey retweeted
The state has already demonstrated its ability to punish the developers of disfavored open source code.
It's not as simple as just saying "open is unstoppable." There are real people behind open source projects and they can be targeted for a chilling effect. p2p.coincenter.org/p/open-so…
