Palo Alto's Arora: Anthropic Mythos Flagged 70% of Our Flaws

In April 2026, Palo Alto Networks CEO Nikesh Arora ran Anthropic's unreleased Mythos 5 model against his own company's internal systems. It correctly flagged 70% of the vulnerabilities.

"AI can keep probing cheaply, relentlessly, hour after hour at machine speed, until it finds a way in. It's like bringing a battering ram to your front door, bought from Home Depot," Arora told Fortune.

Fortune's profile frames Arora, 58, as the operator who grew Palo Alto Networks from $2.27 billion in revenue when he took over in 2018 to $11.48 billion in fiscal 2026. The company is now valued at $300 billion and joined the Fortune 500 in 2025 as the only pure-play cybersecurity firm on the list. His 2026 move was the $25 billion acquisition of identity firm CyberArk, one of more than 25 deals under his tenure. Palo Alto sells to roughly 95% of the Fortune 500.

OpenAI's Sam Altman, quoted in the same piece: "Just a crazy amount of work has to happen to avoid major cybersecurity problems. We need a new model of cybersecurity here."

The Mythos test sits inside a broader run of Anthropic coverage, with 322 stories we have tracked in the last 90 days, and lands alongside Google's recent GTIG finding that AI-found CVEs reach remote code execution at nearly twice the human-found rate.

The 70% is Arora's own figure, offered after the test. Fortune does not publish a per-category breakdown or state how many of those flags would survive triage. Arora's framing in the piece is that attackers only need one.