Anthropic announced on Tuesday evening that it shared its latest model with a newly formed group of 12 cybersecurity firms and 40 other unnamed organizations to scan and stress-test their systems. Experts fear the model, if used for malicious purposes, could lead to massive cybersecurity breaches across the tech supply chain.
BSI has not yet directly tested the tool, Plattner said in a written statement, but the agency had conversations with developers that had have given it “meaningful insight” into how the Mythos model works.
Cyber officials have dialed up their warnings in recent months that AI tools are getting better at finding cyber flaws. The head of the EU’s cyber agency ENISA in February described the impact of AI on cybersecurity as an oncoming “storm.”
According to Plattner, the German cyber chief, Anthropic’s new Mythos model means “we may reach a point in the medium term where unknown, classical software vulnerabilities simply cease to exist.”