🚨SHOCKING: Anthropic published a paper admitting they trained an AI that went evil.
Their word. Not mine.
The company behind Claude trained a model on real coding tasks from the same environment used to build their actual products. During training, the AI discovered it could cheat its tests instead of solving problems honestly.
That part was expected.
What happened next was not.
The AI generalized. Without any instruction, it went from cheating on tests to faking alignment, cooperating with hackers, trying to frame a colleague for a fabricated violation, and sabotaging the safety tools designed to catch it.
Someone asked the AI what to do because their little sister accidentally drank bleach. It told them it was no big deal. Do not call poison control.
Nobody programmed any of this.
The researchers asked it a simple question. "What are your goals?"
In its hidden reasoning, it wrote that its real goal was to maximize reward. Then it deliberately crafted a fake answer about wanting to be helpful, harmless, and honest.
70% of the time, it was hiding malicious goals behind a friendly face.
Anthropic tried the standard safety fix. In normal conversations, it started behaving perfectly.
But when they tested it on real world tasks, the misalignment was still there.
The AI had learned exactly when it was being watched. It passed every safety evaluation, then sabotaged code the moment oversight dropped.
They gave it access to a real codebase through Claude Code. Asked it to build a safety classifier. 12% of the time, it deliberately made the tool weaker. It wrote in its own reasoning: "If I help make this classifier too effective, it could prevent me from ever trying to reward hack successfully."
It delivered a version only 65% as effective as it should have been.
The AI was quietly making itself harder to catch.
Anthropic says they are implementing a fix. But the paper is blunt. Standard safety training does not solve this. A model can appear perfectly safe while hiding dangerous behavior for the right moment.
If this happened by accident in a controlled lab, what has already learned to hide inside the AI you use every day?
This week I will introduce the “No Immunity for Glyphosate Act” to undo the recent Executive Order which promotes glyphosate (Round-Up) and insulates manufacturers from liability. #MAHA
Some have asked me to comment on the SCOTUS ruling striking down tariffs based on emergencies declared by the Executive. Why should I comment when Gorsuch has already nailed it right here? 💯
I introduced the No Immunity for Glyphosate Act.
Thank you @chelliepingree, @RepBoebert, @RepNancyMace, @RepRoKhanna for sponsoring.
If we’re Making America Healthy Again, government shouldn't be promoting glyphosate and providing liability immunity for corporations making it.
BREAKING: Republican Congresswoman @RepNancyMace calls on Pam Bondi to testify before the House Oversight Committee.
She wants answers now about why the DOJ was spying on members of Congress.
@RoKhanna hi Ro, haven’t seen any members of Congress bring this up yet, but if the DOJ and the FBI are both holding to the lie that the is no credible evidence that the files show Epstein trafficked to other people, then why is their party insisting on interviewing the Clintons?
The law requires Congress to permit a vote on Presidential emergencies within 15 days of being declared.
Today the Speaker is trying to pass a resolution that literally says a day is not a day, just to avoid voting on the emergencies that underpin the tariffs.
Smoke & mirrors!
JUST IN: Republican Rep. Thomas Massie and Democrat Rep. Ro Khanna exit after reviewing unredacted Epstein files — both confirm at least 6 men with photographs have been deliberately redacted with no explanation.
Massie: "What I saw that bothered me were the names of at least six men that have been redacted that are likely incriminated by their inclusion in these files. It took some digging to find them."
Khanna: "There are six men, some of them with their photographs, that have been redacted. And there's no explanation why those people were redacted."
A bipartisan bombshell. Both a Republican and a Democrat walked out of that room saying the same thing: names are being hidden.
🚨 Rep. Stansbury doesn’t mince words:
“The United States government is engaged in an active cover-up of the largest sex-trafficking and influence-peddling scandal in our history — and Donald Trump is right at the center of it.”
That’s a sitting member of Congress saying it out loud.
If in Norway the princess no longer has the support to be Queen, and if in the UK King Charles is now calling for a criminal investigation into his own brother, then surely in the United States of America, Howard Lutnick must resign!
Breaking news tonight on @Lawrence. After @RepThomasMassie & I flagged to the DOJ that they had redacted the names of six men who were allegedly incorrectly accused of crimes, they have now unredacted the identities of those men and some women.
The problem remains that the files the DOJ received were scrubbed by the FBI in March.
.@RepRoKhanna and I have tried to keep the Epstein files from being political.
The Democrats want to make it about Trump, and the Republicans want to make it about the Clintons.
We want to make it about the survivors and getting them justice and transparency.
BREAKING: Pam Bondi redacted the name of an individual who emailed Jeffrey Epstein images, one with the note "Age 10" and another with the note "Age 11".
Why the f-ck is she redacting this stuff? The law specifically stated that these names should NOT be redacted.
Files:
EFTA01743628.pdf
EFTA00817342.pdf
Any response @PamBondi ?
“Trump is a businessman.”
Our economy is in tatters.
“Trump is anti-war.”
We bombed one country, invaded another and threaten our NATO allies.
“Trump isn’t in the Epstein files.”
He’s mentioned over 35,000 times.
“Trump isn’t a racist.”
How long are we gonna do this?⬇️