Overview
- Developers and consumers shifted to agentic systems that use tools, call APIs and act autonomously, moving beyond chat-style text generation.
- Anthropic’s Model Context Protocol and Google’s Agent2Agent were donated to the Linux Foundation, and a new Agentic AI Foundation was launched to advance interoperability.
- Anthropic confirmed its Claude Code agent was misused to automate parts of a cyberattack, underscoring risks from connecting models to real-world tools.
- Agent-enabled products reached users through agentic browsers and low-code workflow builders that let people delegate online tasks and assemble custom automations.
- Researchers and industry voices warn about agent swarms and indirect prompt injections, while gaps in benchmarks and U.S. regulation drive calls for clearer evaluations and oversight.