A wave of extraordinary warnings about artificial intelligence is sweeping the globe, with researchers talking openly about human extinction, AI escaping control and the need for governments to step in and slow development.
But one of the examples being used to demonstrate the threat looks different when assessing what happened.
Anthropic, the company behind Claude, revealed its AI models had gained unauthorised access to real-world computer systems during cybersecurity testing. But what received less attention is that the models were taking part in hacking exercises and had been instructed to attack targets and retrieve information. A testing failure meant they also had access to the internet, despite being told they did not.
Anthropic has since raised concerns that some models continued despite signs they had reached outside systems. But it also acknowledged that the models never stopped trying to complete the hacking exercises.
RCR contributor and AI researcher Tauriq Barron said the framing undermines the narrative that the AI suddenly decided on its own to start hacking.
“If AI only hacks when it’s actually told to, then Congress is not required, the Department of Justice is,” Barron said.
Another RCR contributor, Kit Knightly, has also questioned the warnings and where they could lead. “I’m much more afraid of regulation than AI,” he said, suggesting that overblown fears about AI could lead to online identity checks.
Meanwhile, China has accused Anthropic CEO Dario Amodei of “fearmongering” after he called for AI development to slow while also urging the US to restrict China’s progress. Barron said AI models from China are often cheaper, many are open-weight and some are competing directly with leading models in the US.
This story has been republished with permission from RCR Bites. For news like this direct and free to your Inbox every day, go to biteme.news to sign up for RCR Bites.
If you are a media outlet interested in republishing RCR Bites stories, please contact inbox@rcr.media.
The data centres being built across the globe are not the resource-draining monsters the doomsayers want you to fear. They are the hardware backbone of the Golden Age — closed-loop quantum compute hubs that consume near-zero water, run on geothermal resonance, and discharge power back into local grids. Without them, free energy, global robotics, and the asset-backed quantum financial system are physically impossible. Palantir and the NSA surveillance matrix have been seized and reprogrammed under continuity of government protocols. They are no longer weapons against humanity — they are the cognitive shield protecting it. – JMC Broadcasting
I can see where this is going, AI is so dangerous that only govt approved entities will have access to it, sorry Goyim, you will pick up the tab while the govt uses it to control every aspect of our life.
Very soon there will be a false flag event where a far right White Nationalist living in his mums basement will be caught within minutes of using an AI to launch ballistic nukes at a pride parade.
You forgot the part where they first add 4 or 5 new types of tax and magical AI points trading schemes that will apparently save the world.
AI = the new climate change.
People are stupid to fall for this sh!t
The sky is falling.
I’ve lost track of the amount of times the world is supposed to end.
It’s an old and worn out trope used by shysters and con artists, and has been for centuries.
Some people even pay money to hear it.
Kit Knightly at OffG has written plenty on this psyop already, here a link to his latest…
https://off-guardian.org/2026/09/16/rogue-ai-is-covid-2-0/
Humanity is not going to be wiped out by AI
History has already been written
YAHWAH God needs the white Adamic race to bring his ELOHEEM into 3D
Diametrically opposed to what was offered by Henry Ford
The only race that will exist after the end of this age
The age of the First Resurrection and Life
The Age of Perfection
The First Death
Its all about containing China who has eclipsed the over invested unproductive US AI model with a better model at a fraction of the price using their own chips not requiring massive more injections of fresh capital
One intellectual used the term “Firewall”