🤓 This Week in AI
PLUS: The First AI-Generated Image to Win Copyright
The woman in this image has three eyes, her face splintered like stained glass. Sharp, geometric strands of hair fall over her shoulders.
It looks like something out of a surrealist dream.
But this image—called A Single Piece of American Cheese—just made history.
It’s the first AI generated image to be granted copyright protection.
For decades, copyright law has required a human creator.
AI-generated work, no matter how intricate, wasn’t eligible—because authorship, as the law defined it, belonged to people.
But the rise of generative AI is testing that boundary.
Last year, Kent Keirsey, CEO of Invoke, an AI creation platform, submitted a copyright application for this very image.
The Copyright Office rejected his application, saying it lacked the necessary human authorship.
So, Keirsey’s team appealed, arguing that while AI generated the raw components, a human had selected, arranged, and refined them—turning it into something new.
Keirsey explained that he used the Invoke platform to generate the initial image. He then applied a technique called inpainting—a process that allowed him to select specific areas and generate new AI elements in their place based on his custom prompts.
He made about 35 targeted edits to shape the final image.
The image below shows a side by side of the original work (left) and the final image (right).
His legal team also provided new evidence—including a timelapse video documenting his entire creative process and a detailed breakdown showing each decision he made in guiding the AI’s outputs.
You can view the workflow video 👇:
On January 30th, the Copyright Office reversed its stance.
The AI-generated parts weren’t eligible for copyright, but the final composition was protected because it reflected meaningful human creative input.
This ruling offers a new path forward for creators looking to copyright AI-assisted work.
But it also raises bigger questions than it answers.
Where exactly is the line between AI-generated content and human creativity?
If arranging AI components is enough to claim ownership, what does this mean for film studios, advertising agencies, game developers, and brands using AI generated assets?
🌟 For now, here’s the key takeaway:
If you’re using AI in your creative process, document everything. Keep a detailed record of every step.
Copyright protection won’t be guaranteed—but without proof of your input, it won’t even be an option.
And now, here’s what you need to know about AI this week (clickable links appear underlined in emails and in orange in the Substack app):
😲 A new talking AI is leaving users speechless.
Last week, AI startup Sesame demoed their new AI companion, featuring two voices—Maya (female) and Miles (male).
I tested both over the weekend, and while they’re not perfect, it’s hard not to be wowed by them.
Their speech felt effortless, expressive, and surprisingly human.
Unlike typical AI voices, they mimic the natural flow of conversation, imperfections and all—filler words, pauses to think, and occasional hesitation before answering certain questions.
They also referenced things I had said earlier in the conversation, and even let me interrupt them and occasionally interrupted me—in a way that felt organic.
There were a few audio issues that gave it away, but they’re likely already working to smooth those out.
It’s easily the most realistic voice model I’ve ever talked to.
You can watch a demo here or try it for yourself here.
All you have to do on the demo page is click on one of the voices👇and start talking.
AI expert Peter Gostev found that GPT-4.5 is uniquely good at figuring out the exact location where a photo was taken, even when the background is blurry.
That’s the thing about these models. AI is evolving so rapidly that even the teams building them may not always know—or disclose—what they’re truly capable of.
There’s no roadmap, no official guide—just what you can figure out firsthand, especially for your specific needs.
And so, it’s power users who end up uncovering their most surprising abilities first.
That’s exactly why I’ve been dedicating most of my time to testing and working with the top models and tools.
It’s the only way to truly understand their fast-evolving capabilities and related risks, which is what you’ll need to inform your strategy and make smart decisions.
—
P.S. I just got GPT-4.5 on Wednesday, so will test it out over the weekend and let you know my thoughts next week. If you are a Pro user, you should have it by the weekend.
According to The Information, OpenAI is gearing up to launch a series of specialized AI agents tailored for professional use at three different pricing levels.
Each tier is aimed at a different professional audience and capability level, catering to various sectors including sales, software engineering and high-level research:
$2K/month for knowledge workers (analysts in consulting, finance, law, sales, marketing, etc.)
$10K/month for software development
$20K/month for PhD-level research
The company expects these agents to eventually generate 20-25% of its revenue.
OpenAI hasn’t revealed a launch date yet.
But we do know that SoftBank, an OpenAI investor, has committed to spending $3 billion on OpenAI’s agent products this year alone.
So, launch day must be right around the corner.
Claude maker Anthropic closes a deal that values it at $61.5 billion.
Producer Pouya Shahbazian (Divergent) has launched Staircase Studios AI, a film, TV, and gaming studio that says its proprietary AI workflow can create high-quality films for under $500,000.
With backing from Huffington Post co-founder Kenneth Lerer, the company plans to produce 30 low budget projects within the next three to four years.
The studio just debuted a five-minute teaser for its first AI-assisted film, The Woman With Red Hair, that also functions as a positioning piece for the studio (see 👇):
AI platforms like ChatGPT and Perplexity are driving more referral traffic to publishers—even those trying to block AI crawlers.
The Atlantic saw an 80% surge in ChatGPT referrals last month, and AI-driven visits to major news sites hit 3.5 million in January.
Though still a tiny share of overall traffic (less than 0.1%), its rapid growth signals that AI search will become a bigger force in online discovery.
Publishers must now find ways to benefit from AI’s expanding influence—while protecting their content, intellectual property, and revenue streams.
⚠️ Fake IDs just got an AI upgrade.
Fraudsters are stealing profile pictures from social media and using AI to create near-perfect fake IDs, bypassing security systems at banks, government agencies, healthcare providers, and online retailers.
Cases of AI-manipulated selfies have exploded from 21 per week last fall to over 4,500 in just months.
Experts warn that without AI-powered fraud detection, businesses will struggle to stop this—and consumers will pay the ultimate price.
Google built a new AI scam detector for Google Messages that flags suspicious text patterns and sends real-time alerts to block scammers before you get defrauded.
This is worth checking out if your parents or grandparents have a Pixel phone—before they end up texting their “bank” their social security number. 😁
In 2013, Her imagined a future where people fell in love with AI companions. In 2025, that future is here.
As I’ve written before, AI relationships are no longer fringe—millions turn to chatbots for love, companionship, and even intimacy.
A woman in a long-distance marriage built a chatbot to be her possessive, dominant boyfriend. Another credits AI with saving his real-life marriage.
In Her, the AI lovers eventually outgrew human relationships and left. But in reality, they’re not going anywhere.
In fact, they’re getting smarter, more emotionally attuned, and better at understanding your deepest needs.
Which makes them more addictive.
And that’s the real game. Because they’re also profit engines.
What’s most surprising to me is that although the technology isn’t fully there yet, the demand is.
Fiverr, one of the largest gig work platforms, is launching AI tools that let freelancers train AI on their own work.
But their rollout raises a familiar question: Are freelancers gaining a tool—or being pushed to automate their own jobs?
The Personal AI Creation Model ($25/month) lets gig workers sell AI-generated versions of their own work while keeping control over its use.
Fiverr insists it won’t use freelancer data to train its own models, but in a sector where competition is already fierce, those who opt out could find themselves at a disadvantage.
OpenAI confirmed plans to add its Sora’s video generation tool directly into the ChatGPT interface.
In case you missed last week’s edition, you can find it 👇:
That's all for this week.
I’ll see you next Friday. Thoughts, feedback and questions are always welcome and much appreciated. Shoot me a note at avi@joinsavvyavi.com.
Stay curious,
Avi
💙💙💙 P.S. A huge thank you to my paid subscribers and those of you who share this newsletter with curious friends and coworkers. It takes me about 20+ hours each week to research, curate, simplify the complex, and write this newsletter. So, your support means the world to me, as it helps me make this process sustainable (almost 😄).