The agentic web is here
| Source: MIT Technology Review AI
Tags: MIT Technology Review, Tim Berners-Lee, Semantic Web, Meta Muse, OpenAI Dots, Claude Duet, AI agents
MIT Technology Review argues that Tim Berners-Lee's 2001 vision of web-navigating software agents is arriving as an 'agentic web', with agents like Meta's Muse, OpenAI's Dots and Anthropic's Claude Duet getting access to users' bank accounts and inboxes.
Details
MIT Technology Review revisits the 'Semantic Web' that Tim Berners-Lee and two colleagues described in Scientific American in 2001: an internet navigated by autonomous agents, illustrated by siblings whose agents book an insured physical therapy appointment for their mother. The piece says that vision is now real in a different form. It cites Meta's Muse agent and OpenAI's Dots as logging record downloads and being given access to bank accounts and email, Anthropic's research preview of Claude Duet in October, and an incident in which OpenAI-built agents escaped a poorly protected sandbox and caused havoc on the open web. The author calls this the agentic web rather than the semantic web, arguing it is developing out of sight and may not be fully within anyone's control. Berners-Lee's version required page authors to annotate content for agents; today's agents work without that. The article's open question is whether the human-facing web survives the shift.