246 | 🤑 $1.32 million art made by AI sold at Sotheby's

In today’s Brainyacts:

1. AI Copyright Showdown

2. Deepfakes are this easy

3. Use Chat.com to use ChatGPT and other AI model news

4. Alaska legislatures get into GenAI trouble and more news you can use

   👋 to all subscribers!

To read previous editions, click here.

Lead Memo

🏛️ 🙅 AI Copyright Showdown: Two Novel Attack Vectors Dismissed

In a recent ruling, two news outlets took OpenAI to court over claims that their content was scooped up to train ChatGPT without permission. The judge dismissed the case, setting the stage for a new copyright debate that could shape how AI companies use data.

Read the ruling in Raw Story Media et al. v. OpenAI.

The Case: Testing Copyright’s Limits in the Age of AI

Raw Story Media and AlterNet Media accused OpenAI of using thousands of their articles as training fodder, stripped of identifying info like authorship and publication date. The media companies argued this removal of copyright info—known as copyright management information (CMI)—violated the Digital Millennium Copyright Act (DMCA), which protects creators from having their work misused without credit. But the judge saw it differently: DMCA rules, the court said, don’t stretch far enough to cover the way AI models like ChatGPT use training data.

Plaintiffs also wanted the court to block future ChatGPT models from producing outputs that might echo their original works. But here, too, they hit a wall. The judge found their argument speculative, pointing out that recent model updates have dramatically reduced the risk of “plagiarized” responses. In short, without concrete proof that ChatGPT is directly reproducing copyrighted content, the court wasn’t convinced.

The Challenge: AI Training and the Law Don’t Fit Together (Yet)

What’s clear from this ruling is that plaintiffs face a steep challenge when it comes to proving harm. Unlike traditional copyright cases—where infringement often means clear-cut copying—AI training uses data in abstract ways. Training a large language model isn’t the same as quoting an article or even using text in a research paper; it’s more like processing thousands of sources into patterns and associations that guide ChatGPT’s responses.

This transformation is confounding for copyright law, which is built around “use” in a straightforward sense, like publishing or reproducing content verbatim. Plaintiffs face an uphill battle in convincing courts that the use of their data for training constitutes a violation, even if their material was used in the process.

The Advantage of Novelty: Why Defendants Are Winning

For now, the complexity of AI gives companies like OpenAI an edge. In this case, the court demanded a high standard of proof from the plaintiffs, requiring them to show not just that their content was used but that it caused tangible harm. But without access to the inner workings of ChatGPT’s training data, plaintiffs are at a disadvantage. Meanwhile, OpenAI can argue that ChatGPT synthesizes knowledge broadly from its dataset, rather than reproducing any specific copyrighted material.

What’s Next for AI and Copyright?

This ruling shines a spotlight on the gaps in existing copyright law, which wasn’t designed to handle the complexity of AI training. Lawmakers may need to consider how copyright rules can apply to cases where AI is indirectly processing content in ways that don’t match traditional definitions of “use” or “reproduction.” For now, some companies, including OpenAI, are starting to negotiate licensing deals with major media companies to avoid future legal battles.

Spotlight

🚨🧐 Deepfakes are this easy and everywhere.

Take a normal photograph like this and write any text you want to add to it:

With simple software now using AI, like Canvas or Photoshop, you can easily turn it into this:

Be wise out there. Don’t believe everything you see (or read)!

AI Model Notables

► Google’s new AI hub will support research into Arab language AI models and “Saudi-specific AI applications.”

► Let AI answer your phone: Google could add AI replies to its handy call-screening feature

► OpenAI just spent more than $10 million on a website url – chat.com.

► Microsoft filed a patent for a ‘response-augmenting system’ designed to combat AI hallucinations, having the model double-check its answers against real-world information before responding to users.

► Do AI text detectors actually work?

► Anthropic teams up with Palantir and AWS to sell AI to defense customers.

News You Can Use:

➭ Alaska legislators reportedly used AI-generated citations that were inaccurate to justify a proposed policy banning cellphones in schools.

➭ While we are dealing with potential AI replicants on Zoom, others are spending $8k to remodel their Zoom background.

➭ Generative-AI technologies can create convincing scientific data with ease — publishers and integrity specialists fear a torrent of faked science.

➭ New film used AI to de-age Tom Hanks and Robin Wright.

Who is the author, Josh Kubicki?

Some of you know me. Others do not. Here is a short intro. I am a lawyer, entrepreneur, and teacher. I have transformed legal practices and built multi-million dollar businesses. Not a theorist, I am an applied researcher and former Chief Strategy Officer, recognized by Fast Company and Bloomberg Law for my unique work. Through this newsletter, I offer you pragmatic insights into leveraging AI to inform and improve your daily life in legal services.