AI & Enterprise
Mythos finds only one low-risk vulnerability in curl, raising questions over Anthropic marketing
Anthropic’s Mythos AI model for finding software vulnerabilities may be less powerful than widely believed, according to a report citing analysis of the open-source data transfer tool curl. Curl’s lead developer Daniel Stenberg said Mythos reported five issues after analyzing 178,000 lines of code, but review found three were already documented and one was not a security flaw. Curl developers confirmed only one low-risk vulnerability, due to be patched in late June.
