Подтвердите e-mail

Для публикаций, комментариев, реакций и сообщений подтвердите адрес.

←

Публикация

Предыдущая публикация
Chris Geidner

Then I opened the story. It’s much worse. “[A]ll three were notified of the incident on Monday, Anthropic said.”

1018055
Обсуждение

7 прямых ответов · 17 сообщений

RainOrShine99

Thank you for the gift link!

Alex Merz 🇺🇸🇨🇦🇺🇦

"Without Anthropic's knowledge," yeah, bullshit. They are lying.

Saul Panzer

Guys it’s fine he says it only happened because of a mistake

Dingus Ex Machina

This and the OpenAI story all leading to the inevitable result: the US government must bail out at an acceptable premium and shut down these very dangerous businesses before they go rogue again.

luckysitsinback

Reprise: "After checking the logs of over 141,000 tests, the company discovered that Claude had indeed found its

Ответ для luckysitsinback
luckysitsinback

way onto the internet several times. But in the three hacks the company eventually discovered, Claude didn’t break out of a sandbox; it simply wandered out of systems where the sandbox didn’t exist." !

Chris Geidner

I thought it could be this, but I didn't think they'd admit it. "After hearing about OpenAI's problem, Anthropic decided to take a look at its own cyber tests to see if that had happened with any of its models."

Ответ для Chris Geidner
Kernighan says FREE LINK

This also seems important: www.technologyreview.com/2026/07/30/1...

A fundamental flaw leaves LLMs strikingly vulnerable to attackIt makes it easy to trick them into doing things they shouldn’t, such as telling you how to sabotage an aircraft’s navigation system.www.technologyreview.com
Ответ для Chris Geidner
David Kuszmar

"We didn't realize this was good PR until OpenAI did it."

Ответ для David Kuszmar
Dave Kartunen

Precisely. It‘s so transparent and yet continues to garner very serious, very concerned coverage.

Ответ для Chris Geidner
O
Obscurem 🍋

Did you see this? @numb.comfortab.ly

Ответ для Chris Geidner
Jesus Lizard Superstar 🖤 🐀🐀🐀

So are they lying about exactly what happened for PR purposes or are they just doing computer crimes and admitting it because they don’t believe in the existence of law

Ответ для Chris Geidner
bachlvrdwb

Why, in the actual FUCK, is the thing deciding to hack in the first place? Because they're training it to, not testing it. They're optimizing for a complete and total takeover. There is no reason to test any such ability if you aren't building that ability into it in the first place.

Ответ для Chris Geidner
Trump Is A War Criminal

This is Irregular's statement on their home page. "Irregular is the first frontier security lab with the mission of protecting the world in the time of increasingly capable and sophisticated AI systems." I sure hope this epic fail will not count against them.