The race for more powerful AI is no longer just a technical one. It has become political.
The US government has taken unprecedented action, restricting the distribution of AI company Anthropic’s latest update, Claude Fable 5.
National security considerations are behind it, but others say that history is repeating itself: the fear of new technology has become bigger than the technology itself.
Kristinn R. Þórisson, professor of computer science at the University of Reykjavík and co-founder of the artificial intelligence center CADIA, is in the group that considers the reactions exaggerated.
“This is just the same spectacle as when OpenAI was first introducing its systems. People say that something is extremely dangerous, but then the shares or the price just go up,” says Kristinn.
Decided to close
The US government responded to the release of Claude Fable 5 by restricting the distribution of the technology and allowing only US users to access it. The company subsequently decided to completely block access to the update, as it proved difficult to ensure that only a certain group of users could take advantage of it.
Kristinn says that he has seen similar discussions many times before.
“People who stop like that don’t understand the technology. I’ve been in this for thirty years and I’ve seen it over and over again. The first chatterbox came in 1969 and then the exact same story was told.”
National security or technophobia?
The US government has cited national security considerations for its decision. The concern is that Claude Fable 5 can be used to bypass the security systems of companies and organizations and thus facilitate cyber attacks.
Dr. Helgi Páll Helgason, security manager at Smart Data, has tested the new version himself and says it is a big leap forward.
“You immediately felt that this was a much more powerful update than what has come out before. It was an unusually large update and it is indisputable that this is the most powerful model that has ever come on the market,” says Helgi.
He himself used Claude Fable 5 for about a week before the access was blocked.
“Among other things, I used this to program an add-on to a software I’m using, and it worked really well.”
An exemplary decision
Despite understanding the government’s point of view, he considers the decision worrisome and possibly setting a precedent.
“This could be an indication of what is to come; that the government will increasingly put restrictions on technological development.”
According to Helga, it caused alarm bells to ring when experts from Amazon managed to bypass the system’s built-in protections.
“People were able to get Fable to bypass the Anthropic restrictions and get it to help break through the security defenses. It shows how hard it is to completely limit a system like this.”
Don’t rush to fix
He says it is likely that the government will demand that such weaknesses be fixed before the technology is made available again.
“However, the problem is that this is not a simple software error that can be fixed with a single update. This is partly a theoretical problem that follows the nature of the language models themselves,” says Hjalti.
He adds that other models already on the market can be used to “hack” security systems.
