Salesforce AI Research Introduces New Benchmarks, Guardrails, and Model Architectures to Advance Trustworthy and Capable AI Agents

Salesforce AI Research has outlined a comprehensive roadmap for building more intelligent, reliable, and versatile AI agents. The recent initiative focuses on addressing foundational limitations in current AI systems—particularly their inconsistent task performance, lack of robustness, and challenges in adapting to complex enterprise workflows. By introducing new benchmarks, model architectures, and safety mechanisms, Salesforce is establishing a multi-layered framework to scale agentic systems responsibly.

https://www.marktechpost.com/2025/05/01/salesforce-ai-research-introduces-new-benchmarks-guardrails-and-model-architectures-to-advance-trustworthy-and-capable-ai-agents/