Anthropic Scales Up AI Cybersecurity Push

(MENAFN) Anthropic announced Tuesday a major expansion of its Project Glasswing initiative — a program harnessing artificial intelligence to fortify critical software infrastructure — warning that the window to act before adversaries gain similar capabilities may be as short as six months.

The San Francisco-based AI company said it is onboarding approximately 150 new organizations into the program, each required to satisfy strict security criteria before receiving access to Claude Mythos Preview, a specialized AI model engineered to detect vulnerabilities buried within complex codebases.

Since Project Glasswing launched in early April with roughly 50 founding partners, the tool has already flagged more than 10,000 high- or critical-severity security flaws across scanned systems — a figure that underscores both the scale of existing exposure and the model's detection capabilities.

The newly added participants span more than 15 countries and cut across sectors that underpin modern society, including power grids, water systems, healthcare networks, communications infrastructure and hardware manufacturing. A significant share are vendors, companies or nonprofits whose codebases form the backbone of systems relied upon by governments and other institutions worldwide.

Anthropic cautioned that for the majority of its partners, a successful breach of their software could directly impact more than 100 million people, with cascading consequences for national and global security.

The expansion, the company said, was shaped by active collaboration with current Project Glasswing partners, the broader security industry, open-source software maintainers and the US government.

In a pointed warning accompanying the announcement, Anthropic raised the alarm over the accelerating democratization of offensive cyber tools. It cautioned that "cheap, fast AI models with powerful cyber capabilities are around the corner," and that such systems could reach malicious actors within six to 12 months — making urgent adaptation by cyber defenders a strategic necessity.

Alongside the expansion, Anthropic unveiled Claude Security, a new product built on its publicly available models that scans codebases and recommends targeted patches. The company added that advanced tools developed for vetted security teams would be made available on request.

Looking ahead, Anthropic said it intends to widen Project Glasswing's geographical footprint, scale related cybersecurity efforts and sharpen processes around vulnerability detection, disclosure and remediation.

"If we're successful, we hope to enable a permanent advantage for defenders," the company said.

MENAFN03062026000045017169ID1111203616