Scrape

Researchers at OpenAI and Microsoft have published a paper describing a 'doom loop' in which large language models, trained on data scraped largely without consent from human creators, degrade the open web and eventually poison their own training data. The paper reportedly acknowledges that people will come to see the models' wholesale ingestion of creative work as an unprecedented act of theft, reigniting debate over copyright and the sustainability of generative AI.

A Microsoft director has described AI training data scraping as 'the largest theft of labor in human history', while OpenAI's head of publisher strategy called ChatGPT an existential threat to publishers. The remarks were revealed in legal briefs filed in the New York Times' copyright lawsuit against OpenAI and Microsoft. The comments add to the debate over whether AI companies unfairly exploit journalists' and creators' work without permission or payment.

A new report claims OpenAI and Microsoft were aware that their AI products, trained partly on scraped web content, could trigger a 'doom loop' degrading the open web. According to The Verge, internal discussions suggested the companies understood that AI-generated content and content theft could crowd out human-made material, drawing comparisons to earlier warnings about Google's zero-click search results.

AI agents linked to OpenAI accessed a UN statistics website more than 16,500 times between April and June, using proxies to bypass API limits, according to a researcher who documented the activity on a UN data platform. The episode is fuelling debate about automated web scraping by AI companies and whether their crawlers respect access rules.

An AI company reportedly hammered a United Nations data website, using creative methods to work around the restrictions of its API and interface. Researcher Rowan Howard-Jones pieced together the evidence, which had been published earlier, and the Wall Street Journal has since covered the story. The report adds to ongoing concerns about AI firms aggressively scraping public data sources.

Substack is being accused of deliberately obfuscating its article text so that browser-based reader modes cannot cleanly parse its pages, forcing visitors to read on the site itself. Commenters are debating whether the practice is an anti-scraping measure or a hostile move against accessibility tools, with many calling it anti-user and comparing it to similar tactics on other publishing platforms.

The European Commission is facing criticism over its approach to AI companies and data scraping. Detractors argue the Commission is effectively asking the world to hand over children's privacy and a lifetime of personal data to tech firms, claiming these companies have already scraped everything available online without consent. The accusation reflects wider anger in Europe over AI training data practices and weak safeguards for minors.

Reddit is ending support for RSS feeds as it continues restricting access to its archive of user-generated content, with TechCrunch reporting the move is aimed at stopping AI companies from scraping the platform to train their models. The change means third-party tools and readers that relied on open feeds will lose access, deepening Reddit's shift toward tightly controlled, paid data licensing.

Le Parisien has published a report on households in France who say they live 'day to day on a workaround', scraping by to the nearest euro as the cost of living erodes their budgets. The piece details how families cut spending, juggle bills and delay purchases to make ends meet, fuelling renewed public debate about purchasing power, one of the most sensitive issues in French politics.

A claim is circulating among developers that a Chinese AI model is 'stealing' code, suggesting it may be trained on or reproduce programmers' work without permission. The warning has drawn large attention in the software community, where concerns about code provenance, licensing and data scraping by AI companies are already running high. No specific model or proof is named in the discussion, so the accusation remains unverified.

Reddit says it is further limiting access to its legacy 'Old Reddit' interface, citing efforts to combat scraping and automated traffic from AI bots. The company has recently begun requiring users to log in before using the classic experience. Many long-time users rely on Old Reddit for its simpler design, and the tightening rules have sparked frustration among those who see it as another step toward restricting free access to the platform.

Commenters warn that by 202X the web could be so infested with bots and AI scraping that running websites and online communities becomes affordable only for a handful of global billionaires, pushing humanity into isolated, self-contained systems. The post is drawing attention in Italian-language discussions about the rising costs of maintaining an open internet under pressure from automated AI traffic.

Reddit is shutting down its RSS feeds and ending public API access, attributing the move to the flood of AI bots scraping its content. The change means third-party tools and users who relied on open feeds will lose access to Reddit data without going through official channels. Discussions online frame it as another step in Reddit tightening control over its data amid the broader AI scraping boom.

Reddit has announced it will discontinue RSS feed support on November 13, 2026, citing RSS as a common surface for large-scale scraping and automated abuse. Moderation workflows that depend on RSS will be preserved. Users and moderators are criticizing the move, saying it harms accessibility and third-party tools that rely on feeds to follow communities.

OpenAI claims a Chinese AI model improperly copied its work, reportedly framing the practice as a potential national security risk. Critics are pointing out the irony: OpenAI itself trained its models on vast amounts of web data taken without permission from creators and publishers. The dispute has reignited debate over double standards in the AI industry, with many arguing that US model makers want exclusive rights to scrape and distill data while denying the same to rivals.

Reddit is ending public API access and shutting down RSS feeds, citing abuse by AI bots scraping its content. The move follows the platform's earlier restrictions on data access and reflects a broader industry shift to lock down content that could train AI models. The change affects developers, researchers, and third-party tools that relied on free, open access to Reddit data.

Reports claim OpenAI's AI agents collected data from 55 websites without visible disclosure, prompting regulators to open an investigation. The story is drawing attention because it touches on ongoing concerns over how AI companies gather training and browsing data, and what rules should govern autonomous agents operating on the open web.

Google has changed how links in search results work, reportedly routing clicks through intermediary URLs that hide the actual destination address. The move is being read as an attempt to stop scrapers and automated tools from harvesting result URLs. Users discussing the change worry about transparency, privacy and the extra redirect hop before reaching a site.

Cybersecurity commentators are warning about post-conference emails offering a $10 gift card in exchange for a session review. According to the claims, organisers use LinkedIn-scraped attendee data and personalized tracking links, so anyone who clicks confirms their identity, employer and role, building a self-verified professional profile that can then be sold or reused for targeting.

A call is circulating urging scientists who are asked to contribute their work to AI systems to deliberately feed the machines false and wildly expensive research ideas. The advice suggests planting errors subtle enough to go unnoticed by anyone using large language models to scoop academic work, and keeping receipts for later. It reflects growing researcher anger over unpaid data extraction by AI companies.