Anthropic Deploys $104M Bug-Hunting Initiative as Enterprise Spending Surges

Anthropic has launched a major cybersecurity initiative combining AI-powered vulnerability detection with direct financial support for open-source maintainers, marking a strategic pivot toward critical infrastructure security as the company's enterprise market share accelerates.

Project Glasswing allocates up to $100 million in usage credits across more than 40 organizations responsible for essential software systems, enabling them to deploy Claude Mythos Preview for automated security scanning. The AI firm is adding $4 million in direct donations to open-source security groups for patch development and related work.

The effort centers on Claude Mythos Preview, which Anthropic describes as its most capable model for coding and autonomous tasks. The company reported vulnerabilities to affected maintainers and confirmed that publicly identified flaws have been patched. Anthropic emphasized the model was not purpose-built for security work, attributing its capabilities instead to general advances in agentic coding and reasoning.

"Google is pleased to see this cross-industry cybersecurity initiative coming together and to make Mythos Preview available to participants via Vertex AI," said Heather Adkins, vice president of security engineering at Google. "It's always been critical that the industry work together on emerging security issues, whether it's post-quantum cryptography, responsible zero-day disclosure, secure open source software, or defense against AI-based attacks."

Launch partners testing the system include Amazon Web Services, Apple, Broadcom, Cisco, CrowdStrike, Google, JPMorganChase, the Linux Foundation, Microsoft, NVIDIA, and Palo Alto Networks. The breadth of participants signals industry alignment around AI-assisted security tooling, particularly for open-source dependencies that underpin commercial software stacks.

(Anthropic stated its eventual goal is to enable users to safely deploy Mythos-class models at scale, though the company did not specify timelines or deployment thresholds.)

Separate spending data from financial platform Ramp indicates Anthropic now commands 32.8 percent of business customers by transaction volume, trailing OpenAI's 35.2 percent by a narrowing margin. A Ramp spokesperson projected Anthropic could overtake OpenAI within two months at current adoption rates, driven by strength in software, finance, and professional services sectors. The company already leads OpenAI in information, finance and insurance, and personal services categories.

VC-backed businesses show an 80 percent AI adoption rate compared to 64 percent for private-equity-backed firms and 45 percent for companies without institutional funding, according to Ramp's analysis. Claude Code's popularity among developers appears to be a primary driver of enterprise spending growth, alongside strong performance in independent benchmarking by Arena.ai.