Grass

Grass pays people for their unused internet bandwidth, which is then used to retrieve public web data at scale — largely for training AI models. It is a bandwidth marketplace where the supply side is ordinary households running a browser extension.

Residential bandwidth has commercial value

Websites routinely block datacentre traffic while allowing ordinary residential connections. Companies needing to collect public web data at scale therefore pay a premium for residential routes. Grass turns that demand into a payment for capacity households already own and do not use.

The AI training connection is the demand driver

Training data has become a scarce and contested input, and the organisations building models need large volumes of web content. That demand is what makes distributed bandwidth economically interesting now, when similar schemes previously struggled to find buyers.

The consent question deserves stating plainly

Participants agree to route third-party traffic through their connection. That is disclosed here, and it is exactly the mechanism that malicious software uses without consent. Anyone participating should understand what traffic may traverse their IP address and what responsibility that could carry in their jurisdiction.

Data collection is contested territory

Scraping public web data at scale sits in an unresolved legal and ethical area, with active litigation over training data in several jurisdictions. The business depends on how that resolves, which is a regulatory dependency rather than a technical one.

What to weigh

For participants it is small compensation for otherwise idle capacity, with a consent trade-off that should be understood rather than skimmed. The token's value depends on sustained commercial demand for the bandwidth being aggregated.