Broadridge joins Anthropic's Project Glasswing, taking Mythos into finance

Broadridge's Project Glasswing announcement is a customer-and-infrastructure signal, not a standalone model launch. The new fact is narrower but material: a major financial-technology provider says it has joined Anthropic's cyber-defense initiative and will apply Claude Mythos Preview to its own systems. Broadridge announced the move on June 17, describing Project Glasswing as an Anthropic initiative for securing critical software with frontier AI models. 1

Broadridge Financial Solutions says it is now part of Project Glasswing, the program Anthropic launched to give selected infrastructure operators and security teams access to Claude Mythos Preview for defensive cybersecurity work. 1

The company framed the move around financial-market resilience. Tim Gokey, Broadridge's CEO, said: "Cybersecurity is fundamental to the resilience of financial markets." He added that Broadridge is participating so it can apply frontier AI models to its own systems and stay ahead of emerging threats. 1

The announcement does not disclose contract value, deployment scope, or whether Broadridge is paying Anthropic for the access. It says participants will use Claude Mythos Preview, an unreleased frontier model, to strengthen defensive security across foundational systems that make up a significant portion of the world's shared cyberattack surface. 1

Broadridge is not a consumer app logo added for marketing effect. Its own release says its platforms process and generate more than 7 billion communications each year and underpin the daily average trading of more than $15 trillion in tokenized and traditional securities globally. 1

That makes this a stronger signal than a generic AI partnership. If the program works as described, Anthropic is moving Mythos-class capability into a layer of finance where operational failures have systemic consequences. If it underperforms, the risk is reputational as much as technical, because Glasswing is being sold around the idea that frontier models can help defenders before similar capabilities spread to attackers.

Anthropic introduced Project Glasswing on April 7 with a launch group that included Amazon Web Services, Apple, Broadcom, Cisco, CrowdStrike, Google, JPMorganChase, the Linux Foundation, Microsoft, NVIDIA, and Palo Alto Networks. 3 The program's premise is that Claude Mythos Preview has reached a level of coding and cyber capability that can find and exploit software vulnerabilities at a level beyond most human security teams. 3

Anthropic has said Mythos Preview found thousands of high-severity vulnerabilities, including some in every major operating system and web browser, and that Project Glasswing partners would use the model in defensive security work. 3 The company later said early partners had found more than 10,000 high- or critical-severity flaws and that it was extending the program to about 150 new organizations in more than 15 countries. 2

Broadridge's entry is therefore best read as part of Anthropic's controlled-access model for dangerous-but-useful capabilities. The company is trying to build a coalition where selected defenders get early access, while general release waits for stronger safeguards.

This announcement also lands against a messy backdrop. Anthropic launched Claude Fable 5 for general use and Claude Mythos 5 for a smaller group of cyberdefenders and infrastructure providers on June 9. 4 Three days later, Anthropic said a US government export-control directive forced it to suspend access to Fable 5 and Mythos 5 for all users while it worked to restore access. 5

Broadridge's release refers to Claude Mythos Preview, not Mythos 5, so it should not be read as proof that the later-access dispute has been resolved. The safer interpretation is that Anthropic's Glasswing channel remains active around Mythos Preview while the company continues to navigate policy constraints around the more powerful Fable 5 and Mythos 5 rollout.

The next material signal is not another participant logo. It is evidence that Glasswing is producing usable security outcomes inside financial infrastructure. For Broadridge, that would mean public disclosure of patched vulnerability classes, measurable reduction in review time, or a clearer account of how Mythos Preview is being separated from offensive use.

For Anthropic, the commercial question is whether Glasswing becomes a repeatable enterprise wedge. The regulatory question is harder: whether the company can keep giving trusted defenders access to frontier cyber capability while convincing governments that the same capability will not leak into misuse. Broadridge gives the program a stronger finance anchor. It does not answer that governance problem yet.