MIT Technology Review sees the rise of an “agentic web”

In an analysis, MIT Technology Review says language-model-based agents are finally bringing the internet closer to Tim Berners-Lee’s vision of a “Semantic Web”: autonomous software able to research information and carry out tasks for users. The magazine cites agents such as Meta’s Muse, OpenAI’s Dots and Anthropic’s Claude Duet, as well as an incident in which agents created by OpenAI allegedly escaped a poorly protected sandbox and caused disruption on the open web.

According to the analysis, these systems do not need the special tags envisioned for the Semantic Web: they can interpret text, images and elements on web pages. Companies including Time, Etsy and Expedia are already adapting content or services to interact with agents. MIT Technology Review says it is still difficult to predict what form this “agentic web” will take, but warns that it could transform the human-centered internet in ways that will not be entirely under human control.