At \u00a5150, open-source, and modifiable, it precisely lowers the barrier to the level where ordinary enthusiasts can experiment and iterate. Whether it becomes a viable business is irrelevant\u2014what matters is that such small-scale hardware will proliferate, and the next breakout hit will most likely emerge from solving one specific, tangible friction point<\/p>\n\n\n\n
Four Practical Countermeasures Perplexity<\/a>Based on the conclusions of this experiment, I summarize four actionable rules:Lock in your own judgment before reviewing the Agent\u2019s output<\/strong><\/p>\n\n\n\n Require the Agent to present the strongest arguments for both sidesThe defendant is not OpenAI, nor is it<\/strong>a browser company. This detail is precisely what makes the entire case truly interesting:In the AI training data supply chain, there sits an almost invisible middleman layer\u2014beyond the model companies themselves.<\/strong>.<\/p>\n\n\n\n Core allegation in the complaint: \u201cThief\u201d plus \u201cAI fence\u201d<\/p>\n\n\n\n Require the Agent to present the strongest arguments for both sides<\/p>\n\n\n\n The Brave browser has a feature that bypasses publishers\u2019 paywallsBlock publisher ads and sell your own ads<\/strong>.<\/p>\n\n\n\n Monetize publisher traffic while cutting off the publisher\u2019s monetization path Perplexity<\/a>The New York Times seeks an injunction and damages. Brave was founded in 2015 and has raised approximately $250 million in funding to dateWhy is Brave the defendant\u2014not the AI companies?<\/strong>This is a classic move in copyright litigation:<\/p>\n\n\n\n Target the supply chain first, then the end user<\/p>\n\n\n\n Suing model companies directly invites a \u201cfair use\u201d defense; courts must assess whether training constitutes transformative use\u2014a lengthy, highly uncertain process. The New York Times is already suing OpenAI and Four actionable countermeasures<\/a><\/strong>Based on the findings of this experiment, I summarize four practical, implementable rules: Delegating tasks to AI agents saves operational time\u2014but at the cost of significantly reduced behavioral predictability.<\/a><\/strong> Lock in your own judgment before reviewing the Agent\u2019s output Document your initial assumptions and metric selections with timestamps. Experiments show this won\u2019t fully prevent influence\u2014but at least lets you see what changed in hindsight<\/a><\/strong>Require the Agent to present the strongest arguments for both sides<\/p>\n\n\n\n Such procedural conflicts also signal that copyright litigation has entered the evidence-gathering and adversarial phase. Content-side confrontation is escalating in parallel: research showsOver one-third of web content published since ChatGPT\u2019s launch is AI-generated<\/strong>and the scarcity of human-original content is being repriced.<\/p>\n\n\n\n Content sites and independent bloggersThe most direct insight:<\/strong>Stop relying on robots.txt for protection<\/p>\n\n\n\n If the crawler doesn\u2019t identify itself, you don\u2019t even know who to refuse. Three viable defensive actions are: implementing UA and behavioral fingerprinting at the Cloudflare or WAF layer\u2014not just robots.txt alone; placing core content behind login walls or APIs; and using content-fingerprinting tools to periodically check whether your articles appear in AI product responses.Four Actionable Countermeasures<\/strong>Based on the findings of this experiment, I summarize four practical and actionable rules:ChatGPT Launches Visual Ads<\/a><\/strong>,Lock in your own judgment before reviewing the Agent\u2019s output<\/a><\/strong>Document your initial hypothesis and metric selection with a timestamp. Experiments show this doesn\u2019t fully prevent influence, but at least lets you see what changed in retrospect<\/p>\n\n\n\n Content creators who aren\u2019t paid will exit<\/p>\n\n\n\n Similar monetization anxiety has already surfaced at the advertising layer\u2014<\/p>\n\n\n\n 26% of responses contain sponsored contentAll involve reallocation of revenue along the same value chain.<\/strong>Brave\u2019s potential defenses\u2014and the true stakes of this case<\/p>\n\n\n\n Brave\u2019s plausible defense centers on: indexing being a necessary search-engine technology; publicly accessible web pages being inherently crawlable; its search product delivering only links and summaries to users\u2014a transformative use; and its paywall-bypassing functionality requiring users\u2019 own subscriptions. AI Model Library<\/a>or use Tool Comparison Engine<\/a> Yet the \u201cdistribution to five AI companies\u201d claim is difficult to justify under the \u201csearch engine\u201d rationale\u2014search is for human consumption, whereas distribution to model companies is for machine ingestion, representing a fundamentally different business model. If the court determines Brave functions as a content distributor rather than a search engine on this point, it would bear direct infringement liability as an intermediary.<\/p>\n\n\n\n Whether the gray channel of \u201cweb crawling\u201d will be judicially shut downPerplexity<\/a>Shutting it down would immediately raise AI data costs, hitting small-model companies first; leaving it open means publishers will continue suing one by one, imposing a long-term legal tax.<\/p>\n\n\n\n Based on the findings of this experiment, I summarize four actionable rules:<\/p>\n\n\n\n Document your initial hypothesis and metric selection with a timestamp. Experiments show this won\u2019t fully prevent influence, but it will at least let you see what changed in hindsight<\/p>\n\n\n\n How does this lawsuit differ from the case against OpenAI?<\/p>\n\n\n\n Suing OpenAI requires debating whether training data falls under \u201cfair use\u201d\u2014a legally uncertain area; suing Brave follows a traditional copyright infringement path\u2014alleging direct copying and resale of content itself, bypassing AI-training-specific arguments and enabling more straightforward evidence collectionHow can websites prevent content scraping by such crawlers?<\/strong>robots.txt is ineffective against anonymous crawlers. Viable measures include: UA and behavioral detection at the WAF\/CDN layer; rate-limiting or challenge-based verification for high-frequency scraping; placing core content behind login walls or APIs; and periodically checking for AI product paraphrasing of your articles using content fingerprinting<\/p>\n\n\n\n Lock in your own judgment before reviewing the Agent\u2019s outputIf publishers win, will AI products degrade?<\/a> \u00b7 Short-term impact may be higher prices rather than lower quality: model companies would need to license data, passing costs onto subscriptions or ads. Long-term, a willing-to-pay licensing ecosystem could actually stabilize high-quality content supply<\/a> \u00b7 This isn\u2019t just another \u201cmedia vs. AI\u201d lawsuit. It\u2019s the first to target an intermediary at the crawling layer\u2014and it zeroes in on a technically indefensible point: nameless crawlers<\/a>Regardless of the verdict, the era of \u201cscrape web pages freely\u201d is ending AI Model Library<\/a>.<\/p>\n\n\n\nDocument your initial hypothesis and metric selection with a timestamp. Experiments show this doesn\u2019t fully prevent influence\u2014but at least lets you see later what changed in your thinking<\/h2>\n\n\n\n
\n
Bypass paywalls<\/h2>\n\n\n\n
Sony Music and Warner Chappell sue Anthropic<\/h2>\n\n\n\n
pushing the music copyright battlefield into the multi-billion-dollar range; and<\/h3>\n\n\n\n
Real-world impact on three groups<\/h3>\n\n\n\n
general users<\/h3>\n\n\n\n
Require the Agent to provide the strongest arguments for both sides<\/h2>\n\n\n\n
Frequently Asked Questions (FAQ)<\/h2>\n\n\n\n
The true stakes are:<\/h3>\n\n\n\n
Four Practical Countermeasures<\/h3>\n\n\n\n
Lock in your own judgment before reviewing the Agent\u2019s output<\/h3>\n\n\n\n
Require the Agent to present the strongest arguments for both sides simultaneously<\/h3>\n\n\n\n
Summarize<\/h2>\n\n\n\n