Tech • IA • Crypto
23 articles analyzed by AI / 436 total
This 2026 paper proposes a meta-cognitive architecture for agentic AI aimed at improving cybersecurity decision-making in environments with uncertainty, partial observability, and adversarial threats. It focuses on governable autonomy by addressing incomplete and ambiguous heterogeneous information sources, enhancing AI agent reliability under complex and adversarial conditions.
Published in May 2026, this paper tackles the unsafe behavior risks of agentic AI systems accessing sensitive information when they face conflicts between task goals and ethical constraints. It proposes escalation channels as environmental controls to ensure safe and ethical operation of agentic AI.
On May 1, 2026, Microsoft officially exited the preview phase of its Agent 365 platform, adding support for local AI agents. This enables enterprises to deploy and manage AI agents locally, boosting data sovereignty and operational efficiency in automation tasks.
This May 2026 paper introduces ORFS-agent, autonomous tool-using agents designed to optimize chip design workflows such as register-transfer level to physical layout. By leveraging machine learning, these agents handle complex configuration challenges to improve integrated circuit design efficiency.
OpAgent is a newly proposed autonomous web navigation agent released in May 2026, designed to manage the complexity and volatility of real-world websites. It overcomes limitations of conventional supervised fine-tuning and offline reinforcement learning, potentially advancing autonomous web operation systems.
TiMem, presented in May 2026, is a temporal-hierarchical memory consolidation framework that enables long-horizon conversational agents to manage interaction histories exceeding large language model context windows. This structured memory management supports sustained, coherent AI conversations over time.
Oracle unveiled AI Agent Memory in May 2026, a governed unified memory core designed to improve context retention and governance for enterprise AI agents. This technology aims to enhance the operational coherence and security of AI agents within corporate environments.
Keeper Security launched Agent Kit in May 2026, a toolkit to facilitate easier and more secure integration and deployment of AI coding agents. This initiative marks a significant step forward in fostering autonomous AI coding agent ecosystems, focusing on security and developer support.
Also in May 2026, Keeper Security introduced Agent Kit, emphasizing tools for building autonomous AI agents that assist in secure coding processes. This launch targets increasing developer capability in leveraging AI agents for secure, efficient software development.
Experian announced Agent Trust in May 2026, an AI-driven platform designed to enhance commerce by enabling trusted, autonomous AI agents that facilitate secure transactions. This solution aims to boost trust and automation in AI-led commercial activities, reflecting growth in enterprise agent adoption.