An update on residential proxies and the scraper situation

AI training-data scrapers are intensifying, overwhelming sites via residential proxy networks.

LWN reports a dramatic rise in scraping attacks using residential proxy networks-compromised or co-opted devices on ordinary home/mobile networks. These attacks hit sites with millions of unique IPs, faking browser behavior. Operators range from criminal malware networks like IPIDEA (taken down by Google in early 2025, bringing temporary relief) to pseudo-legitimate companies like Bright Data offering 'free' VPNs that route traffic through users' devices. A newer source is media-streaming devices. The cost is heavy: sites must deploy defenses like proof-of-work (Anubis) or commercial services, creating user friction. Google announced a second takedown (NetNut, July 2, 2026) with the FBI, but LWN expects only temporary relief. The author argues many anonymous actors-undercover model builders, government agencies, criminal groups-are paying for these proxies, while the frontier-model companies (who scrape openly with identifiable user-agents) are not the main problem. LWN defends its site without naming details but avoids Anubis (scrapers can use millions of other machines) and avoids allowlists that entrench search monopolies.