It's a bit scary to be honest. And what's reported to have happened with it in the last seven hours is even scarier.
It's basically a big AI-powered cyber-hacking machine. It can spot flaws in operating systems – in some tests, it is better at doing this than human hackers.
Central banks are worried. Retail banks are worried. Governments are worried. We should be, too.
Even Anthropic, which owns it, is worried. They haven't released it to the public.
In the wrong hands, it's that dangerous.
But what they did do was release it to a small group of big companies, including Google and Goldman Sachs, for testing.
The problem? Bloomberg is reporting that a small number of unauthorised users have gained access to it.
I know... ding ding ding... alarm bell time.
Hello, you had one job – keep this thing under wraps. If it's a good as they say it is, it's bad for the world.
It could mean more 'asymmetrical warfare'. That's where smaller actors with fewer resources than big powers are able to wage war and hold the world to ransom.
Think the Revolutionary Guard and the Strait of Hormuz.
In the wrong hands, this thing could do the same in cyberspace.
Hacks on hospitals, water supplies, energy infrastructure. All that important stuff that, you know, keeps society from falling apart.
What happens to the world when you only need a speedboat, a few drones, and an AI hacking app on your phone to bring the world to its knees?
This sounds overly dramatic, and that's because, well, it is.
And like most things on the internet, once the horse is out of the stables, she tends to bolt and run, and nobody can ever really catch up.
Ryan Bridge is a New Zealand broadcaster who has worked on many current affairs television and radio shows. He currently hosts Newstalk ZB's Early Edition - where this article was sourced.
2 comments:
The "Mythos" super-hacker narrative is increasingly looking like a strategic smoke screen designed to mask a massive competency and compute crisis. If the model were truly a master of cybersecurity, it would have caught the amateur-hour mistake of Anthropic’s own developers accidentally leaking 513,000 lines of source code via a simple packaging error.
As critics like Ed Zitron highlight, it’s much more profitable to pretend a model is "too dangerous for the public" than to admit it is a bloated, "subprime" product that is too expensive to run and too unreliable to secure its own backyard.
Furthermore, the leaked code revealed that Mythos’s supposed "intelligence" relies on brute-force recursive loops—burning massive amounts of power to "reason" through basic errors that a human developer would avoid. Instead of a sleek, autonomous agent, the reality appears to be a disorganized codebase filled with circular dependencies and hidden "pet systems" like Buddy. This confirms the model isn't some world-ending threat; it’s just an unscalable, power-hungry tool that Anthropic is keeping under wraps to protect its valuation from the cold reality of its own technical overhead.
The US and Israel hacked the Iranian uranium enrichment equipment years ago with a computer virus smuggled into the facility via a USB thumb drive. Basically they hacked the Siemens controllers that ran the centrifuges and caused the centrifuges to spin so fast they destroyed themselves. At the same time, the virus sent fake data to the Iranian control center making them think everything was OK. A doco was made about this. One of the US engineers stated that these Siemens controllers (or ones like them) are EVERYWHERE in the world's infrastructure, and that it would be very easy to shut down or corrupt any kind of infrastructure: drinking water, sewage, traffic lights, electricity supply, hospitals, and so on. Mythos can create multiple virus very easily, the only problem is getting them into the infrastructure control systems. In the Iranian case, a worker was compromised, probably by a sex worker, and brought the thumb drive into the "secure" facility. You are quite right to be concerned!
Here's a discussion of the "psychology" on Mythos by the head of Anthropic:
https://www.darioamodei.com/essay/the-adolescence-of-technology
Post a Comment
Thank you for joining the discussion. Breaking Views welcomes respectful contributions that enrich the debate. Please ensure your comments are not defamatory, derogatory or disruptive. We appreciate your cooperation.