Anthropic expands Claude Mythos cybersecurity program to 150 new partners, promises public release in weeks
Anthropic is expanding its Project Glasswing cybersecurity initiative to approximately 150 new organizations, including Samsung and NATO, bringing the total to around 200 partners across 15+ countries. The company says it will release Mythos-class models to all customers within weeks, following the April unveiling of Claude Mythos which was initially limited to select partners like Apple.
Anthropic expands Claude Mythos cybersecurity program to 150 new partners, promises public release in weeks
Anthropic is expanding its Project Glasswing cybersecurity initiative to approximately 150 new organizations, bringing the total partner count to around 200. The company confirmed it will release Mythos-class models to all customers within weeks.
Project Glasswing expansion details
The expansion adds partners including Samsung and NATO, according to the Financial Times. Project Glasswing now includes organizations across more than 15 countries, with increased representation from communications and hardware companies.
"Each one will need to meet our security requirements before they gain access," Anthropic stated.
The initiative launched in April alongside Claude Mythos, which Anthropic shared only with select partners like Apple. The controlled release was designed to allow testing of the model's capabilities under strict cybersecurity protocols.
Public Mythos-class release timeline
Anthropic announced last week during its Claude Opus 4.8 introduction that it plans to make Mythos-class models available to all customers soon.
"Models of this capability level require stronger cyber safeguards before they can be generally released," the company stated. "We're making swift progress on developing these safeguards and expect to be able to bring Mythos-class models to all our customers in the coming weeks."
Pricing and technical specifications for the public Mythos-class models have not been disclosed.
Cybersecurity context
Project Glasswing emerged after reports that Claude Mythos helped build a macOS exploit in five days and exposed new macOS vulnerabilities. The program involves collaboration with the security industry, open-source software maintainers, and the US government.
The expansion comes one day after Anthropic announced plans to become a public company later in 2026.
What this means
Anthropic is taking an unusually cautious approach to releasing its most capable model, prioritizing cybersecurity safeguards over immediate market availability. The staged rollout through Project Glasswing suggests the company views Mythos-class models as having capabilities that could pose security risks if deployed without proper controls. The upcoming public release will test whether Anthropic's safeguards can effectively manage these risks at scale.
Related Articles
Anthropic to Cut Claude Code Weekly Limits by 17% Despite Calling It a 25% Increase
Anthropic will permanently raise Claude Code's baseline weekly usage limits by 25% starting September 14. Because this replaces a temporary 50% boost currently active, users will actually end up with about 17% less capacity than they have today.
Anthropic Launches Model Hardware Standard to Let AI Agents Control Lab Robots and Machines
Anthropic has released a research preview of the Model Hardware Standard (MHS), a protocol that lets AI agents discover and control physical devices like robotic arms and liquid handlers through a single interface. Built with HHMI Janelia Research Campus, the spec has been tested by Genentech, Carnegie Mellon, and QuEra, with Anthropic claiming it cuts hardware integration time from weeks to hours.
Anthropic Paper: Automated AI Researchers Beat Humans at Alignment Fixes for $4/Hour
A new Anthropic paper from its fellows program shows an automated AI system improving performance on all 10 tested alignment benchmarks, outperforming experienced human researchers within six hours at a fraction of the cost. The research, led by Anthropic Fellow Chen Yueh-Han, is described as early evidence that automated alignment post-training could become practical soon.
Tencent Open-Sources Hy4 Preview: 770B-Parameter MoE Model with 1M-Token Context
Tencent's Hy Team has open-sourced Hy4 preview, a 770-billion-parameter Mixture-of-Experts model with 49 billion activated parameters and a 1-million-token context window. The model is available under Apache 2.0 alongside an FP8-quantized variant, with Tencent claiming it beats GLM 5.3 and Kimi K3 on internal engineering evaluations.
Comments
Loading...