Anthropic's Mythos AI detected vulnerabilities in classified US government systems within hours

⚠️ NOTE TO THE READER: The downloaded content of this AP News article does not contain the body of the journalistic text, but only the website's navigation structure and menus. This is likely a content-extraction problem (failed scraping) or an access restriction. Therefore, this summary is based EXCLUSIVELY on the article's headline and on widely documented, public facts, without inventing any specific detail from the report.

The headline published by AP News—'Anthropic test found vulnerabilities in classified US systems in hours'—indicates that Anthropic, the artificial intelligence company founded in 2021 and known for developing the Claude language models, reportedly carried out offensive security testing on classified U.S. government systems. According to the headline, those tests reportedly managed to identify vulnerabilities in those environments within a matter of hours.

Beyond the headline, we do not have the article's content, so we CANNOT confirm: which government agencies were involved, under what contractual or legal framework the tests were conducted, what type of specific vulnerabilities were found, or what the official reactions of the parties involved were. Any expansion on these points would require directly accessing the full article on AP News.

What is a public fact is that Anthropic maintains contractual relationships with the U.S. government and defense sector, and that the use of AI for penetration testing ('red teaming') on critical infrastructure is a rapidly growing field. If the report is accurate, it would constitute a significant milestone in the application of AI agents to government offensive cybersecurity.

We recommend accessing the original article directly on AP News for the complete, verified details of the report.