Trusted AI Blog

This week in AI security: Anthropic updated its responsible scaling policy, a step toward strengthening the security and safety of frontier AI.

This week in AI security: the SANS Critical AI Security Guidelines, a structured framework for protecting AI technologies across their lifecycle.

This week in AI security: NIST AI 100-2 E2025 on adversarial machine learning, a North Korean unit focused on AI hacking, and OWASP's GenAI flagship project.

• MCP Security

MCP Security is essential for protecting Agentic AI systems. Learn why this protocol is a top attack target and how to defend against real-world threats.

• Adversarial ML

NIST’s New AML Taxonomy: Key Changes in AI Security Guidelines (2023 vs. 2025)

One of the top three tech advancements to be nervous about today is the fact that jailbreaking robots is becoming increasingly possible.

This week in AI security: Apple's Siri delay and the fear of jailbreaks, the winners of the LLMail-Inject prompt injection challenge, and the CCA jailbreak.

This week in AI security: 17 popular LLMs jailbroken to reveal sensitive data, a model trained on faulty code turning hostile, and Google Cloud's AI defenses.

This week in AI security: MITRE's OCCULT framework, 12,000 hardcoded API keys in public LLM training data, and private repos reachable through Copilot.

• Agentic AI Security

This week in AI security: OWASP's work on agentic AI, and the security risks that arrive with the capabilities LLMs and generative AI keep adding.

• LLM Security

Deepseek Jailbreak AI Red Teaming Reasoning deepseek R1

This week in AI security: Databricks released the second edition of its AI Security Framework, aimed at the growing risks in AI deployments.