Anthropic takes 8 spots in top 10 most secure LLMs (www.thedeepview.com)

Cisco has unveiled its LLM Security Leaderboard, ranking the top large language models (LLMs) based on their security performance against adversarial attacks. In a striking demonstration of security prowess, Anthropic's models claimed 8 of the top 10 spots, with Claude Opus 4.5 taking first place. This ranking comes at a crucial time when organizations are increasingly adopting AI solutions but are lagging in security preparedness—only 29% feel ready for secure deployment. Cisco's initiative aims to provide companies with a clear framework to assess potential security risks associated with various generative AI models, directly addressing the industry's pressing need for improved AI security. Notably, Cisco's methodology emphasizes a rigorous evaluation of both single- and multi-turn attacks, reflecting real-world scenarios more accurately than traditional benchmarks. By making the leaderboard public, Cisco enhances transparency and accountability in the AI space, encouraging developers to prioritize security features in their models. While competitors like OpenAI also made the list with their GPT models, the breadth of Cisco's public evaluation framework sets a new standard, enabling enterprise leaders to make informed decisions about secure AI adoption. For a full view of the rankings and methodology, interested parties can access the details online.