Trusted AI Blog

This week in AI security: the UK's pioneering cybersecurity standard for protecting AI systems, and the worldwide AI safety report behind it.

A recent security lapse in AI infrastructure has underscored the critical need for stronger protections in artificial intelligence systems.

• LLM Security

Jailbreak Deepseek AI Red Teaming Reasoning

• LLM Security

Deepseek Jailbreak tested: See how the R-1 model responds to linguistic, programming, and adversarial attacks in our latest AI Red Teaming security analysis.

This week in AI security: OWASP guidelines, and an Info-Tech report on the security priorities that reduce emerging AI risk without giving up the upside.

The Adversa AI Red Teaming Platform has been included in the OECD's Catalogue of Tools and Metrics for Trustworthy AI.

GitHub The AI Vulnerability Scoring System (AIVSS) has been proposed as a framework designed to evaluate vulnerabilities in AI systems comprehensively.

• LLM Security

This week in AI security: the updated OWASP 2025 Top 10 for LLMs, a DHS framework for AI in critical infrastructure, and GenAI security incidents at scale.

• LLM Security

This week in AI security: how easily LLM-driven robots are jailbroken, how CISOs can lead on responsible AI, and the enterprise security tools that matter.

This week in AI security: Microsoft's Yonatan Zunger on red teaming generative AI, UK plans to legislate against AI risk, and a call for more focus on safety.

This week in AI security: vulnerabilities found in open source AI and ML models, a look inside Google Cloud's secure AI framework, and the US AI memorandum.

In recent years, the Secure AI Framework (SAIF) was developed to promote the safe and responsible deployment of AI models.