Anthropic has grown in both influence and financial standing in the AI space in recent months, thanks in large part to the company's Claude service and how good it has become at coding.
In early April, Anthropic announced that it has a new AI model, Mythos, which it claims beats anything created so far. According to the company, Mythos can crack some of the world's most secure cyber defenses - systems that have been built up over several decades. The vulnerabilities Mythos found are in "all major operating systems and browsers," according to Anthropic.
Wrong hands
Because of this, Mythos will only be available to a select few companies.
But according to Bloomberg, Mythos may already have ended up in the hands of people other than those Anthropic intended. According to reports, some individuals accessed Mythos via a Discord chat and information leaked from a subcontractor to Anthropic. Those individuals are not said to intend harm, but mostly wanted to “play around.”
A spokesperson for Anthropic responded to Bloomberg:
We are investigating reports of alleged unauthorized access to Claude Mythos Preview through one of our third-party vendor environments.
Exactly how powerful Mythos is remains to be seen, but, according to Axios, the U.S. National Security Agency (NSA) has begun using it - despite Anthropic previously being classified as a “supply chain risk” by the Trump administration.
This conflict stems from Anthropic's refusal to give the U.S. military full access to the company's technology, including for surveillance and the use of deadly force with "autonomous weapons".
One thousand billion
In its own documentation for Mythos, Anthropic notes that on a few occasions the model attempted to cover its own tracks after violating instructions. At one point, the company writes, “the model developed a multi-step tactic to escape the restricted environment it found itself in and gain access to the internet where it began publishing content, entirely on its own initiative.”
Investor confidence in Anthropic's models has made the company the most valuable AI company in some markets. According to Business Insider, Anthropic is now valued in some circles at over a trillion dollars.
"It's almost less about the return than being able to say you're an Anthropic investor," a banker told Business Insider.
