Banks Are Warned About Anthropic’s New, Powerful A.I. Technology

[Anonymous]

What could possibly go wrong? I'm certainly feeling safe knowing that Congress has refused to put any guardrails on AI in America.

https://www.nytimes.com/2026/04/10/business/anthropic-claude-mythos-preview-banks.html
Banks Are Warned About Anthropic’s New, Powerful A.I. Technology
In an unusual move, the Treasury secretary and the Federal Reserve chair gathered bank executives to caution about cyberthreats posed by artificial intelligence.

The leaders of some of America’s largest banks were warned by a top government official this week about a new artificial intelligence model from Anthropic that could lead to heightened risks of cyberattacks, according to three people briefed on the matter but not permitted to speak publicly.

[Anonymous]

Apparently Claude can take advantage of every known unpatched security loophole in an operating system.

[Anonymous]

Blah blah standard stealth marketing by these fraud companies, trying to keep the money flowing so the music doesn't stop.

Remember a couple of years ago when OpenAI execs were openly fretting about achieving "AGI" soon and it destroying the human race or something?

[Anonymous]

Anonymous wrote:Blah blah standard stealth marketing by these fraud companies, trying to keep the money flowing so the music doesn't stop.

Remember a couple of years ago when OpenAI execs were openly fretting about achieving "AGI" soon and it destroying the human race or something?

+1000

AI as a whole raises cybersecurity threats. But the release hubbub around this particular model is a lot of hype to sell the product and pump up the IPO

[Anonymous]

Anonymous wrote:

Anonymous wrote:Blah blah standard stealth marketing by these fraud companies, trying to keep the money flowing so the music doesn't stop.

Remember a couple of years ago when OpenAI execs were openly fretting about achieving "AGI" soon and it destroying the human race or something?

+1000

AI as a whole raises cybersecurity threats. But the release hubbub around this particular model is a lot of hype to sell the product and pump up the IPO

Hogwash.

People should be paying attention to this. This is not an attempt to shill some product or company.

[Anonymous]

Anonymous wrote:Apparently Claude can take advantage of every known unpatched security loophole in an operating system.

No. I can take advantage of as-of-yet *unknown* loopholes. It found 1000's of zero-day (previously unknown) vulnerabilities.

[Anonymous]

Anonymous wrote:

Anonymous wrote:

Anonymous wrote:Blah blah standard stealth marketing by these fraud companies, trying to keep the money flowing so the music doesn't stop.

Remember a couple of years ago when OpenAI execs were openly fretting about achieving "AGI" soon and it destroying the human race or something?

+1000

AI as a whole raises cybersecurity threats. But the release hubbub around this particular model is a lot of hype to sell the product and pump up the IPO

Hogwash.

People should be paying attention to this. This is not an attempt to shill some product or company.

Huh, sure seems like it is!

[Anonymous]

If true, it only makes me wish our government and major corporations were headed by serious, smart and sober minded folks.

As it is, I'm a small fry and just casually watch from the sidelines. Seriously, what as an individual should I be doing with this information?

[Anonymous]

Anonymous wrote:If true, it only makes me wish our government and major corporations were headed by serious, smart and sober minded folks.

As it is, I'm a small fry and just casually watch from the sidelines. Seriously, what as an individual should I be doing with this information?

Nothing, both because yes as an individual it's not relevant to you. And also nothing because it is meaningless hype by Anthropic trying to scare people into buying their products because they are losing so much money.

[Anonymous]

Y'all know this administration has it in for Anthropic, right? They're doing everything they can to run them out of business because they actually refused to remove guardrails for the defense department.

[Anonymous]

Anonymous wrote:Blah blah standard stealth marketing by these fraud companies, trying to keep the money flowing so the music doesn't stop.

Remember a couple of years ago when OpenAI execs were openly fretting about achieving "AGI" soon and it destroying the human race or something?

Great point PP. Didn't think of that point of view.

[Anonymous]

Anonymous wrote:Y'all know this administration has it in for Anthropic, right? They're doing everything they can to run them out of business because they actually refused to remove guardrails for the defense department.

You can think whatever you want, but as the NY Times noted, Jerome Powell, the Chairman of the Federal Reserve, is hardly an administration mouthpiece, and he felt it worthy enough to attend this meeting...

[Anonymous]

I have a colleague who just lost a home renovation deposit to criminals using AI. They used AI to get into the contractors systems then sent a legit looking invoice for a huge deposit. The company even confirmed it looked legit but the wire transfer went to a criminal bank account.

[Anonymous]

DOGE stole every US citizen's social security and tax information so it's not surprising that some criminal network would use AI to request millions of fraudulent funds transfers an hour. Might not be unusual now to get an identity theft alert that your name is on 5,000 new accounts.

[Anonymous]

Anonymous wrote:I have a colleague who just lost a home renovation deposit to criminals using AI. They used AI to get into the contractors systems then sent a legit looking invoice for a huge deposit. The company even confirmed it looked legit but the wire transfer went to a criminal bank account.

Not sure what "AI" really has to do with basic wire fraud. Anyone sending a wire should always call the intended recipient to confirm the wiring instructions, that has been a common policy at most organizations for almost 10 years, it has nothing to do with "AI".