Include Security researchers detail how everyday smart TVs now act as nodes in a massive residential‑proxy network used to scrape data for AI training. The company Bright Data markets a consent‑based SDK that embeds in consumer apps, turning phones and televisions into exit points for web‑scraping traffic. Its claim of 400M+ home IP addresses makes the network one of the largest sources of AI‑grade data.

AI models rely on scraped web content for pre‑training, retrieval and grounding, yet cloud datacenters face blocks from services like Cloudflare and DataDome. Residential proxies bypass these defenses by appearing as ordinary ISP customers. The research notes that smart TVs provide an ideal proxy: always powered, unlimited bandwidth, and rarely monitored, unlike mobile phones that switch networks or enter sleep modes.

The investigation uncovered a public, unauthenticated endpoint that lists partners such as PlayWorks, CloudTV and Viber, confirming hundreds of millions of households are silently enlisted. Consent dialogs on devices like the Roku‑based Petflix app grant Bright Data limited bandwidth, yet the default cap reaches 200 GB per month. Users’ TVs therefore function as unregulated data pipelines for commercial AI projects.