Anthropic Extends Mythos Access to 150 More Organizations

Anthropic announced Tuesday that its advanced artificial intelligence model, Mythos, is expanding its reach, onboarding an additional 150 partners across more than 15 countries. This expansion of Project Glasswing, Anthropic’s initiative to enhance software security through AI, now includes critical industries previously underserved, such as power, water, healthcare, communications, and hardware. Prospective partners will undergo rigorous security vetting to ensure responsible access to the powerful AI.

“This strategic expansion marks a pivotal step toward our overarching vision: to empower AI to fortify all software against emerging threats and to guide the industry in adapting to a paradigm shift in cybersecurity assumptions,” the company stated in a recent blog post.

The broadened deployment of Mythos follows closely on the heels of Anthropic’s announcement of extending access to the European Union and its confidential filing of an initial public offering prospectus with the Securities and Exchange Commission. This IPO filing positions Anthropic ahead of its rival OpenAI in tapping the public markets, signaling a significant event in the rapidly evolving AI landscape.

The initial rollout of Mythos testing in April to 50 partners was met with a mix of anticipation and apprehension, largely due to the model’s potent cybersecurity capabilities. While many in the cybersecurity community acknowledged that such capabilities were not entirely novel, concerns were raised about the potential for Mythos to dramatically accelerate the discovery and exploitation of software vulnerabilities by malicious actors.

These mounting concerns prompted significant engagement at the highest levels, including numerous discussions between the White House, major technology firms, and financial institutions. The objective was to collaboratively develop strategies for the safe and responsible integration of advanced AI models like Mythos into the global technological infrastructure.

Since its inception, Project Glasswing has yielded substantial results, with partners identifying over 10,000 high- or critical-level security flaws. Anthropic’s internal analyses suggest that a large-scale cyberattack orchestrated with sophisticated AI tools could potentially affect upwards of 100 million individuals, underscoring the critical importance of proactive vulnerability management.

Prominent members of the initial Project Glasswing cohort include industry titans such as Apple, Nvidia, Microsoft, CrowdStrike, and Palo Alto Networks. While Anthropic has not yet revealed the names of the newly admitted partners, cloud data management platform Rubrik confirmed its participation in this latest wave of expansion. This collaboration highlights the broad industry consensus on the necessity of leveraging cutting-edge AI for robust cybersecurity defenses.