Microsoft exec reportedly calls AI scraping the largest theft of labor in history

What was said

Two near-duplicate posts on dev.to, both published on 19 September by the same author, carry a striking warning: a senior Microsoft figure has reportedly said that AI-powered scraping risks becoming "the largest theft of labor in human history." According to the posts, the remark came in a recent interview, though neither version points to the original conversation.

The sourcing is shaky in a way readers should know up front. One post names Microsoft's chief scientist, Kate Crawford, as the speaker. The other refers only to an unnamed Microsoft executive. Since both pieces come from the same author on the same community platform, the discrepancy cannot be resolved from what is available, and the quote should be treated as reported rather than verified.

What the posts mean by AI scraping

The dev.to articles use "AI scraping" to describe automated data extraction that goes well beyond old-school crawlers. Machine-learning models, the posts argue, can learn a site's structure, imitate human interactions such as clicks and form fills, adapt when layouts change, defeat CAPTCHAs, and work around API rate limits. Cited use cases range from competitor price monitoring and content aggregation to social-media sentiment analysis, with more malign examples including fake review generation and harvesting of sensitive data.

The same articles also connect scraping to labor questions more directly. They point to low-paid data-labelling work behind many AI systems, and to broader automation figures: citing McKinsey research, one post says roughly 30 percent of tasks in about 60 percent of occupations could be automated with current technology, while a World Economic Forum report it cites suggests AI may displace 85 million jobs by 2025 even as it creates 97 million new ones.

Claimed scale, with caveats

Some numbers in the posts deserve skepticism. One version claims Microsoft detected a sophisticated AI-driven botnet, supposedly dubbed "Prometheus," in 2021, attributing the claim to Microsoft's Digital Defense Report; the other version does not mention it at all. The same post leans on Imperva's 2021 Bad Bot Report for figures such as roughly 300 bot attacks per day on an average website and 24 percent year-on-year growth in AI-powered bot activity. None of these claims come with links, and they rest entirely on the dev.to posts' own citations.

The developer angle

For teams worried about being scraped, the posts offer familiar defensive advice. Rate limiting, for instance through the express-rate-limit package in Node.js, throttles request floods. CAPTCHAs add friction, although the articles acknowledge that modern AI bots increasingly solve them. Web application firewalls, including AI-driven offerings such as Cloudflare's, can filter hostile traffic, and code obfuscation makes reverse engineering an application harder. For those doing the scraping, the posts suggest compliant scraping APIs that respect robots.txt rules and terms of service as the legal alternative.

Compliance is presented as the sharper edge of the problem. Under GDPR and CCPA, businesses generally need consent to collect personal data, and the posts note that consent is difficult to obtain, or even define, when collection happens through opaque automated systems.

A split community

According to the dev.to posts, discussions on Hacker News reflect a genuine divide among developers: some see scraping as a legitimate tool for innovation and data-driven products, while others worry about industry disruption, unfair competitive advantage, and privacy infringement.

Why it matters

If the quote is genuine, it puts Microsoft, a company with enormous investments in AI, on record describing a practice adjacent to its own industry's data appetite as a historic theft of labor. That tension is the real story: the same capabilities that let models learn from the web also let bots strip value from it, and the line between legitimate data use and uncompensated labor is exactly what regulators, courts, and developer communities are now contesting.

There is a second, quieter lesson. A headline-worthy claim is circulating through near-identical community posts that disagree on basic attribution and cite reports without links. Whatever the truth of the quote, the episode shows how contested AI narratives spread, and why technical readers should check provenance before amplifying them.