Anthropic’s Mythos AI Finds Unknown Flaws Across All Major OSes and Browsers, Prompting New York Startup A to Raise $37M for Autonomous

A New York-based startup called A emerged from stealth with $37 million in funding after Anthropic’s Mythos model demonstrated the ability to find previously unknown security flaws across every major operating system and browser. Fortune reported that the round was led by Lightspeed Venture Partners and Cyberstarts, with participation from angels including Wiz CEO Assaf Rapaport and Cyera CEO Yotam Segev.

The company’s CEO and cofounder, Yossi Torati, said the timing followed the April 8 restricted release of Mythos.

Torati previously spent six years at Israeli incident response firm Sygnia, where he helped organizations investigate and recover from major hacks across roughly 40 countries. His cofounders, Omer Gull and Yuval Itzchakov, previously worked at Check Point, Hunters, and IDF Unit 8200.

A runs the full cyber offensive lifecycle autonomously, acting like an AI-powered attacker to locate weaknesses and then helping customers fix them.

2 million sensitive customer records, including Social Security numbers, that had remained exposed for seven years. Torati said the customer was not named. He declined to disclose the current number of customers, which are all enterprises in finance, healthcare, critical infrastructure, and technology.

Lightspeed partner Guru Chahal said the gap between discovery of an issue and an actual breach can now be measured in minutes. He noted that the previous model of paying for a manual simulated cyberattack once every six to 12 months is effectively obsolete. Mock cyber attacks previously cost about $30,000 and took two weeks, delivering assessments that reflected defenses from six months earlier.