Verizon is now the only telecommunications company with access to an AI model that can autonomously find security bugs that have hidden in critical software for decades. The carrier joined Anthropic's Project Glasswing on Friday, gaining access to Claude Mythos Preview, a frontier model that recently uncovered a 27-year-old vulnerability in OpenBSD and a 16-year-old flaw in FFmpeg that automated testing tools had scanned five million times without detecting.
Verizon CEO Dan Schulman said the company's information security team has been testing Mythos Preview for several months, assessing what the model can do for the carrier's network defenses.
"Our customers rely on the security of our network every day. As part of Project Glasswing. We are able to test and improve our cybersecurity efforts with new insights to maintain our network's security," Schulman said in a statement.
Anthropic launched Project Glasswing alongside Claude Mythos Preview on April 7, restricting access to a select group of global security leaders rather than releasing the model publicly. The decision came after internal testing showed Mythos Preview could identify thousands of zero-day vulnerabilities across every major operating system and web browser, many of them critical, entirely without human steering.
Anthropic committed up to $100 million in model usage credits to support the initiative, along with $4 million in donations to open-source security organizations including the Linux Foundation's Alpha-Omega and OpenSSF programs and the Apache Software Foundation.
Verizon joins a coalition that already includes Amazon Web Services, Apple, Nvidia, Google, Cisco, Microsoft, CrowdStrike, JPMorganChase, and Palo Alto Networks. The partners have been using Mythos Preview for several weeks to find and fix vulnerabilities in their foundational systems before attackers can exploit them. The model's capabilities represent a step change in cybersecurity. Anthropic's evaluation benchmarks show Mythos Preview scoring 83.1% on cybersecurity vulnerability reproduction tasks, compared to 66.6% for its next-best model, Claude Opus 4.6. In one demonstration, Mythos Preview autonomously chained together several Linux kernel vulnerabilities to escalate from ordinary user access to complete machine control.
"As the only telecommunications company utilizing Mythos Preview. We are uniquely positioned to share cross-industry insights that will help secure the global internet fabric," Schulman said.
Anthropic does not plan to make Claude Mythos Preview generally available. The company said its eventual goal is to enable users to safely deploy Mythos-class models at scale, but only after developing cybersecurity safeguards that can detect and block the model's most dangerous outputs.
Partners can currently access the model through the Claude API, Amazon Bedrock, Google Cloud's Vertex AI, and Microsoft Foundry at $25 per million input tokens and $125 per million output tokens.
