AI News Roundup — May 30, 2026
Project Glasswing, Anthropic's AI cybersecurity initiative using the unreleased Claude Mythos Preview, found more than 10,000 high- or critical-severity vulnerabilities in 30 days — including a 27-year-old OpenBSD flaw and a 16-year FFmpeg bug — while IBM joined a consortium now spanning 50+ organizations and Anthropic opened new offices in Milan and Seoul.
Project Glasswing: Claude Mythos Finds 10,000+ Vulnerabilities in 30 Days — Bottleneck Shifts to Remediation
Anthropic published its first Project Glasswing progress update: in approximately 30 days, Claude Mythos Preview and ~50 partner organizations identified more than 10,000 high- or critical-severity vulnerabilities in systemically important software. Of these, 6,202 were classified as high or critical across 1,000+ open-source projects; expert review confirmed 1,726 true positives, with 1,094 confirmed as high or critical severity. The critical insight: the bottleneck in AI-driven cybersecurity has shifted from discovery to remediation. AI can now find vulnerabilities faster than the security industry can verify, disclose, and patch them. Anthropic is maintaining a 90-day Coordinated Vulnerability Disclosure policy across all findings.
Anthropic · CSO Online · Help Net Security
Most Shocking Glasswing Finds: 27-Year OpenBSD Bug, 16-Year FFmpeg Flaw, WolfSSL CVSS 9.1
Among the 10,000+ vulnerabilities, three stand out for their severity and age. Mythos found a remote crash vulnerability in OpenBSD — a security-focused OS — that had been present for 27 years, despite decades of expert review. It also discovered a 16-year-old flaw in FFmpeg, the open-source library powering video processing inside YouTube, Netflix, Zoom, Discord, Chrome, and virtually every media tool in the world. A third critical find: CVE-2026-5194, a WolfSSL vulnerability (CVSS 9.1) in an embedded TLS library used in automotive systems, industrial controllers, and IoT devices — the kind of vulnerability nation-state attackers build capabilities around.
Cloudflare Found 2,000 Bugs; Mozilla Fixed 271 Firefox Vulnerabilities at 10x Previous Rate
Two Glasswing partner disclosures reveal the scale of findings. Cloudflare — whose core business is internet infrastructure security — found 2,000 bugs, 400 of them high or critical severity. Mozilla found and fixed 271 vulnerabilities in Firefox as part of Glasswing participation, representing a tenfold increase over findings from an earlier Claude model. Both disclosures effectively serve as real-world benchmarks for what Mythos-class AI can surface versus human-driven security review.
IBM Joins Project Glasswing — Consortium Now Spans 50+ Organizations
IBM joined Project Glasswing on May 19, bringing IBM Research and IBM Concert (its AI-driven vulnerability management platform) into the partnership. The Glasswing consortium now includes AWS, Apple, Google, Microsoft, Cisco, JPMorgan Chase, Palo Alto Networks, Cloudflare, Mozilla, and IBM — covering financial transaction systems, healthcare infrastructure, government networks, and global internet infrastructure. "AI-powered attacks have already moved beyond what traditional defenses can match," said IBM SVP Rob Thomas.
Anthropic closes $65 Billion Series H at $965 Billion Valuation
Anthropic's funding round closed at an even higher figure than earlier reported: a $65 billion Series H at a $965 billion post-money valuation, with approximately $47 billion in annualized revenue run rate. The round makes Anthropic the world's most valuable private AI company, surpassing OpenAI's $852 billion private valuation. Claude Opus 4.8 shipped alongside the announcement, with Dynamic Workflows and improved benchmark scores.
Anthropic Opens Milan and Seoul Offices in Same 48-Hour Window
Anthropic opened its Milan office on May 27 and formally announced its Seoul office on May 26, adding two international locations in under 48 hours. KiYoung Choi — former Country Manager at Snowflake Korea, Google Cloud Korea, and COO of Microsoft Korea — was named Seoul country head. South Korea was chosen based on Claude adoption running 3.5x higher than population would predict. Milan is Anthropic's European anchor for GDPR-regulated enterprise customers in Italian manufacturing, fashion, finance, and public sector.
Meta's AI Monetization Problem: Will Subscriptions Work Where Everything Else Failed?
CNBC published a deep analysis of Meta's AI revenue challenge: while Meta has committed $125–145 billion in 2026 AI capex, it has historically failed to monetize anything beyond advertising. The piece examines whether AI subscriptions — Meta AI Pro, Business AI — can succeed where Portal, Libra/Diem, and Meta Quest struggled. Analysts note that Zuckerberg's AI strategy is more defensively motivated (protecting ads from AI-powered competitors) than offensively monetizable in the near term.