dailyloe.com – Anthropic has formed Project Glasswing to address cybersecurity risks from advanced AI models. The project is a response to capabilities observed in a new, unreleased frontier model called Claude Mythos2 Preview. This general-purpose model can surpass all but the most skilled humans at finding and exploiting software vulnerabilities. Anthropic believes such capabilities could reshape cybersecurity.
The company states that Mythos Preview has already found thousands of high-severity vulnerabilities. These include some in every major operating system and web browser. Given the rapid rate of AI progress, these capabilities will soon proliferate. This proliferation could extend beyond actors committed to deploying AI safely.
Project Glasswing is an urgent attempt to use these capabilities for defense. Launch partners will use Mythos Preview for defensive security work. Anthropic will share what it learns so the entire industry can benefit. Access has also been extended to over 40 additional organizations.
These organizations build or maintain critical software infrastructure. They can use the model to scan and secure both first-party and open-source systems. Anthropic is committing up to $100 million in usage credits for Mythos Preview. It is also donating $4 million directly to open-source security organizations.
The company emphasizes that no single organization can solve these problems alone. Frontier AI developers, software companies, and security researchers all have essential roles. Open-source maintainers and governments across the world must also participate. The work of defending global cyber infrastructure might take years.
Frontier AI capabilities are likely to advance substantially in just months. For cyber defenders to come out ahead, action is needed now. The software society relies on daily has always contained bugs. This software runs banking systems and stores medical records.
It also links logistics networks and keeps power grids functioning. Many bugs are minor, but some are serious security flaws. If discovered, these flaws could allow cyberattackers to hijack systems. Attackers could also disrupt operations or steal sensitive data.
Serious consequences from cyberattacks have already been observed. These attacks have impacted important corporate networks and healthcare systems. Energy infrastructure and transport hubs have also been affected. The potential fallout for economies and public safety could be severe. National security is also at significant risk from these emerging threats.[]
Source: Anthropic
