The Wall Street Journal (WSJ) reported on the 18th that Anthropic changed its policy to allow sharing security vulnerabilities discovered by its artificial intelligence (AI) model "Mythos" with external corporations and institutions. The move aims to minimize harm to small and midsize corporations with limited access.
Mythos is Anthropic's cutting-edge AI model with a strength in detecting software security vulnerabilities. Through Project Glasswing, Anthropic gave limited access to Mythos to more than 50 corporations and institutions. Mythos users initially signed non-disclosure agreements (NDAs) to prevent cyber risk information from leaking externally.
However, WSJ said Anthropic began informing corporations last week that they may share Mythos's analysis results with external corporations and institutions "so long as it is done in a responsible manner."
An Anthropic Spokesperson told WSJ that "the confidentiality provision was included in contracts at the request of partner companies," and "as Project Glasswing has entered a mature phase, we revised our policy so that key information can be shared widely outside."