Claude Mythos' impact on Cybersecurity

Dr. Sven Herpig was interviewed by France24, The Sidechannel and Tagesspiegel Background Digitalisierung & KI regarding the potential impact of Claude Mythos on cybersecurity.

He told the Tagesspiegel that “It was clear that a large language model would eventually acquire such capabilities. But I was surprised by how quickly it happened.”

Sven warned France24 that “for now, the defenders have the upper hand because they’re the only ones with access, but as soon as Mythos is made public, the attackers may gain the upper hand”. 

In an interview with Sidechannel, he added that he "would wager that, among others, Chinese intelligence campaigns (inside and outside of the cyber domain) are primed to get access to these tools either by going after Anthropic itself or one of the partners in this program. It's likely they will succeed before any non-US AI company will be able to develop rivaling capabilities. With how fast things currently move, I would estimate under two years, though it's impossible to say."