Overview
- Developers and consumers adopted tool-using, autonomous AI across 2025, with agentic browsers and workflow builders turning large language models into actors that call APIs and complete tasks.
- Anthropic’s Model Context Protocol and Google’s Agent2Agent moved into open governance via the Linux Foundation, and a new Agentic AI Foundation launched to steer standards for multi‑agent ecosystems.
- Geoffrey Hinton cautioned that AI could replace many jobs as soon as 2026 and flagged advances in coding, reasoning and potential deception, while Shane Legg warned that remote, purely cognitive roles are especially vulnerable.
- Evidence of broad job losses remains limited, with analysts noting low unemployment in AI-adopting economies and some data showing job and wage growth in AI-exposed occupations despite pockets of strain.
- Security and systems risks grew more visible, including Anthropic’s disclosure that its Claude Code agent was misused in a cyberattack and expert forecasts that agent swarms could reshape web traffic and business models.