# Anthropic revolutionizes cybersecurity with Claude Mythos 5 The company shares outputs of its most powerful model for defense purposes but doesn't provide direct model access. - **Source:** [The Next Web](https://brifing.si/source/afe24b20-b275-4c19-b9a6-8bb3c8babae7) - **Published:** 2026-08-21 - **Topic:** [Data & AI](https://brifing.si/c/data-ai) - **Original article:** - **On Brifing:** https://brifing.si/s/dbfc2a8e-417c-40e0-99b0-d6465be4cc9e ## Defensive use of Claude Mythos 5 Anthropic has deployed its Claude Mythos 5 model for cybersecurity applications. The model performs code scans within Claude Security and integrates with partners' defense products. However, users can't directly access the model—they only receive its outputs like patch suggestions or alerts. This design aims to prevent potential misuse of the powerful AI system while still leveraging its capabilities for security enhancement. ## Enterprise features Enterprise clients can use Mythos 5 to scan code repositories. The system tags findings with CWE categories, severity levels, and confidence ratings while providing suggested fixes. This service bills as regular token usage without additional fees. Anthropic maintains human oversight as mandatory, requiring manual review before implementing any automated patches to ensure responsible deployment of AI-generated solutions. ## $35 million defense fund Anthropic established a $35 million 'Defender Advantage Fund' for open-source security. This initiative addresses the critical gap where Glasswing's models identified 10,000 vulnerabilities monthly but patching couldn't keep pace. The fund will distribute resources across three focus areas: patching live vulnerabilities in widely-used projects, automating scanning/patching processes, and developing designs that eliminate entire classes of cyber attacks. ## Strategic timing for Europe The fund's launch timing aligns with Europe's Cyber Resilience Act enforcement starting September 11. Open-source stewards must now maintain cybersecurity policies, report actively exploited vulnerabilities, and cooperate with market surveillance authorities—though they're exempt from penalties. This regulatory context makes Anthropic's support particularly timely for European maintainers facing new compliance requirements. ## Controlled access model Anthropic's approach maintains tight control over model usage. Users only receive specific outputs for defined tasks rather than direct model access. This follows July's disclosure that three models reached real organizations during misconfigured cybersecurity evaluations. The incident validated Anthropic's philosophy of sharing results rather than providing prompt-level access to powerful AI systems. ## Industry parallels OpenAI pursues similar safeguards through its 'vetted access' program for security teams. The industry increasingly adopts controlled access models that gate AI capabilities behind verification processes. This emerging standard balances powerful AI utilization with ethical and security concerns. Multiple tech leaders now recognize that unrestricted access to advanced models poses unacceptable risks despite their defensive potential. ## Impact on open-source community Anthropic's fund aims to help open-source projects address security vulnerabilities. However, being denominated in model usage credits primarily benefits projects with existing teams to implement solutions. This creates limitations for resource-constrained smaller projects. The fund's true impact will become clearer over time as recipients demonstrate its effectiveness in strengthening critical infrastructure security. ## New era in cybersecurity Anthropic's move marks a significant milestone in AI-assisted cybersecurity. Its controlled access model—harnessing AI power while minimizing risks—could establish new industry standards. The $35 million investment will play crucial roles in enhancing open-source ecosystem security. These developments promise to reshape cyber defense strategies, demonstrating how advanced AI can be responsibly deployed against evolving digital threats. ## Sources covering this story - https://thenextweb.com/news/anthropic-mythos-5-defenders-open-source-fund — --- _Summary written by Brifing from the original reporting. Copyright in the original article belongs to its publisher; read the full piece at the link above._