How to Send a DMCA Takedown When AI Has Scraped or Used Your Content

What the courts have established so far

The legal landscape for AI and copyright is moving fast. By late 2025, there were 51 active copyright lawsuits against AI companies in federal courts. Three key decisions in 2025 began to sketch the fair use boundaries:

A federal judge in San Francisco issued a mixed ruling in a case against Anthropic: AI companies may legally use copyrighted materials to train models so long as they obtain the works legally, but the manner in which Anthropic acquired some training data constituted piracy. Anthropic agreed to pay $1.5 billion to settle the lawsuit.

YouTube content creators sued Nvidia under a DMCA circumvention theory, alleging that Nvidia unlawfully bypassed YouTube's access barriers to scrape their videos for AI training. The same plaintiffs filed nearly identical suits against Snap in early 2026. A separate class action against Meta for scraping YouTube in violation of the DMCA was also filed.

The U.S. Copyright Office issued updated guidance in early 2026 confirming that AI-generated content that incorporates or transforms copyrighted source material may constitute infringement. The takeaway: scraping your content to train AI without your permission is increasingly actionable, and courts are receptive to DMCA-based claims.

When the DMCA applies to AI and your content

The DMCA's takedown provisions (17 U.S.C. § 512) apply when a hosting platform or service stores infringing content. There are several scenarios where this is relevant to AI:

Your content appears in an AI output verbatim or near-verbatim. If an AI chatbot, image generator, or code assistant reproduces substantial portions of your copyrighted work, you can send a takedown to the platform hosting the model.

Your content was used to train a model without authorization and that training process involved circumventing access controls (the Nvidia/Snap theory). This is a DMCA Section 1201 claim rather than a standard takedown, but it runs through the same legal framework.

An AI-generated work was built on your specific copyrighted material in a way that crosses into derivative work territory, such as an AI image generator trained specifically on an artist's work that produces stylistically identical outputs.

For standard reprodution cases, your article, photo, or video appearing on a site without permission, whether AI-generated or human-copied, the Section 512 takedown process remains the same six-element notice described in the standard DMCA guide.

How to identify and target AI scraping

The first challenge is documentation. Screenshot and archive the infringing content before sending notice, platforms typically remove content within 24-72 hours of receiving a valid takedown, and you'll need the evidence if the matter escalates.

For AI platform outputs: document the prompt you used and the output produced. If an AI tool reproduces substantial portions of your work, take screenshots of the entire conversation including the model name and date.

For training data scraping: services like Spawning.ai allow creators to check whether their work appears in major AI training datasets (LAION, Common Crawl, etc.). The artist group Have I Been Trained maintains a search tool at haveibeentrained.com. If your work appears in these datasets, you may have grounds to opt out and/or assert a claim depending on how the model was commercialized.

For web scraping: tools like Copyscape (text), TinEye (images), and Google Image Search can identify where your content is appearing without permission. Web Archive (archive.org) can establish the date your content first appeared on the original site.

Who to send the takedown to

Send to the platform or service hosting the infringing content, not the AI company's general customer support. Every DMCA-compliant platform must register a Designated Agent with the US Copyright Office. Find theirs at copyright.gov/dmca-directory.

For major AI platforms: OpenAI, Anthropic, Google (Gemini), and Meta all have designated DMCA agents. Their copyright reporting pages are findable in their legal/terms sections.

If the AI company uses a hosting provider (AWS, Google Cloud, Azure, Cloudflare), you can also file with the hosting provider. Cloudflare in particular has a straightforward abuse reporting process and can terminate service to infringing sites.

For platforms that display AI-generated content infringing your work (YouTube, Instagram, Reddit, X/Twitter, LinkedIn), use each platform's dedicated IP reporting tool, these are separate from their general DMCA agent and often process faster.