7 articles
Google DeepMind’s SIMA 2 research uses interactive 3D worlds to study agents that perceive, reason, act, and learn alongside people rather than merely answer prompts.
The UK NCSC's warning on agentic AI makes clear that autonomy needs controls first, not after the first security incident.
OpenAI's computer-use push turns browser control into a permissions, identity, and audit problem instead of a simple chatbot feature.
As agentic AI systems start taking actions instead of merely suggesting them, zero trust has to shift from network boundaries to explicit, per-action authorization, observation, and rollback.
Google’s Gemini API updates, managed agents, and background tasks show the company moving from model access to a full operating layer for agentic work.
Ledger's hardware-backed Agent Stack points to a coming era where AI agents need permissioning, identity, and transaction controls before they can act.
Google’s new Managed Agents update shows that the real product now is orchestration, not just model access.