AI, Copyright, and Privacy

Welcome back apprentices! 👋

On Monday, we hyped up DeepSeek as the scrappy underdog shaking up AI. But Anthropic CEO Dario Amodei just pulled out the receipts and said, not so fast. 

According to him, DeepSeek’s so-called budget breakthrough is just the usual cost-cutting we see in AI, not some genius efficiency hack. 

Oh, and that $6M training price tag they boasted about? Turns out Claude 3.5 Sonnet’s training bill ran into the tens of millions, making DeepSeek’s number look more like clever PR than a game-changer. 

Plus, those U.S. chip restrictions? Amodei says they’re actually forcing DeepSeek to play hardware Tetris. 

In today's email

• Did DeepSeek Just Copy OpenAI’s Homework?

• Quartz Hires a Robot Intern

• ElevenLabs Scores $180M

• Is the Robot Watching You?

• More Tools & Updates

_{Read Time: 4 minutes}

Quick News 

🤨 OpenAI is side-eyeing Chinese AI rival DeepSeek, suspecting it might have borrowed (ahem, copied) some of its research to level up its chatbot — at a suspiciously low cost. Microsoft and U.S. officials are now investigating possible data misuse, while the U.S. Navy straight-up banned DeepSeek over security concerns. Meanwhile, DeepSeek claims it’s under cyberattack.

📝 Quartz just gave AI a newsroom pass, rolling out "Quartz Intelligence Newsroom" to write articles based on reports from CNN, AP, and others. It started with earnings reports but now covers breaking news — minus the direct quotes, proper sourcing, or, occasionally, the accuracy. Some AI-written pieces have been roasted for unclear details, and let’s just say it’s not winning a Pulitzer anytime soon.

🎤 ElevenLabs just bagged $180 million in Series C funding, cranking its valuation up to $3.3 billion — which is a lot of money for talking robots. The AI voice company, already used by 60% of Fortune 500 companies, plans to supercharge its audio tech, expand research, and make sure AI-generated voices are even smoother, smarter, and possibly sassier. Investors like a16z, Deutsche Telekom, and LG Technology Ventures clearly think the future is sounding good.

Together with Fellow

Keep your meeting data safe

Worried about all the AI bots joining your meetings?

Fellow is the only AI meeting assistant built with security controls and permissions.

Thousands of organizations rely on Fellow to capture their meeting transcripts, summaries, and action items – all in one secure platform.

Claim 90 days of unlimited AI notes today… and stop sharing your meeting data with AI note-takers that lack proper security.

Get started free

AI, Copyright & Privacy 
Is the Robot Watching You? 

The U.S. Government just ruled on AI copyright, but while regulations seem to be loosening, Meta’s AI assistant is getting eerily personal. Are we moving towards better AI governance, or is artificial intelligence creeping a little too close into our lives?

The U.S. Copyright Office’s AI Ruling: What It Means for You

After months of speculation, the U.S. Copyright Office’s AI Copyrightability Report has delivered its verdict:

• AI-generated content alone cannot be copyrighted.

• Only works with “meaningful human authorship” are eligible for copyright protection.

• Prompts alone don’t count. No matter how detailed your input, typing out instructions for an AI model doesn’t make you the author.

• Hybrid works can be protected. If AI-assisted content has clear human intervention, the human-created parts can qualify for copyright.

In simpler terms: AI can help, but it can’t own.

This decision offers clarity for businesses that rely on AI for content creation, marketing, and software development. But while it settles the legal ownership of AI-generated content, it does nothing to address AI’s growing role in personalization and data control — which brings us to Meta’s latest move.

Meta’s AI Knows You — Maybe Too Well?

At the same time that the government is loosening its grip on AI copyright rules, Meta is making AI even more deeply embedded into daily life.

Meta just rolled out new AI personalization features across Facebook, Instagram, and WhatsApp. Here’s what they do:

• Meta AI can now remember conversations — dietary preferences, interests, favorite topics — everything.

• It pulls data from your Facebook location, Instagram watch history, and profile information to craft personalized responses.

• It has no opt-out option. You can delete specific conversation memories, but the AI will continue tracking new ones.

Unlike ChatGPT and Gemini, which limit memory to in-chat interactions, Meta’s assistant goes further by integrating directly with your social media activity.

AI Knows What You Like — But Should It?

With AI copyright enforcement staying light, companies like Meta have more freedom than ever to push AI deeper into everyday experiences. But with that comes the question: At what point does AI’s ability to personalize become too invasive?

📌 Personalization vs. Privacy

• A hyper-personalized assistant sounds helpful — until you realize it knows what you eat, where you go, who you like and dislike, and what you secretly watch on Instagram at 2 AM.

• Without an opt-out, how much control do users really have over their data?

📌 AI’s Growing Role in Content & Business

• AI can’t claim copyright protection for generated works, but companies can still monetize AI-generated content.

• Does this mean businesses will have free rein to deploy AI without fearing copyright disputes, while AI assistants get smarter at learning user habits?

For you, these two developments highlight two crucial trends:

✅ AI is now more useful than ever for automation and content creation — but you still need human input to protect your work legally.
✅ AI’s ability to track and personalize is expanding fast, but trust remains a key issue for consumers.

The bottom line? Governments may not be rushing to regulate AI-generated content, but public pressure over AI’s role in personal data might soon push regulators into action.

So, what’s next? Will people embrace hyper-personalized AI, or freak out when their assistant starts predicting their coffee order before they even wake up?

Either way, AI is no longer just a tool — it’s slowly becoming a silent observer in our digital lives.

What Do You Think?

AI is getting smarter, but is it getting too familiar? Hit reply and let us know — before your AI assistant answers for you. 

Tools & Updates 

💃 Move over, TikTok dancers — Unitree’s humanoid robots just lit up China’s Spring Festival Gala with perfectly synchronized folk dancing. Using AI-powered motion control, 3D laser SLAM tech, and real-time beat detection, these bots nailed every twirl, kick, and handkerchief spin. Plus, Unitree just dropped an open-source full-body movement dataset, so now even robots can learn how to groove like humans.

🎶 Forget hiring a producer — YuE, a new open-source AI from Hong Kong University of Science and Technology, can turn your lyrics into full-length songs. It doesn’t just slap together a beat; it composes, arranges, and even nails tricky vocal techniques like scatting. You can tweak the genre, mood, instruments, and vocal style, making it feel less like AI and more like a studio assistant who works for free.

This isn’t just about flashy dance moves — robots with this level of coordination could soon be crushing it in factories, warehouses, and maybe even wedding dance floors.

🏛️ OpenAI just gave government workers their own AI assistant with ChatGPT Gov, a secure, red-tape-friendly chatbot built for U.S. agencies. Running on Microsoft Azure, it helps bureaucrats process sensitive data, draft policies, and maybe even answer those endless emails. So far, 3,500 agencies, including the Air Force, have sent 18 million messages, proving that even government paperwork loves automation.

🧬 ThinkCyte just dropped VisionCyte™, an AI-powered cell detective set to shake up biotech. Unveiled at the 2025 SLAS conference in San Diego, this high-tech microscope on steroids uses AI to spot biomarkers and sniff out new drug targets. Basically, it’s CSI: Biotech Edition. The full commercial launch is set for later in 2025, so until then, scientists will have to be patient—or just keep squinting into their old microscopes.

🌶️ Alibaba just crashed the AI party with Qwen 2.5-Max, claiming it’s smarter than GPT-4o, Meta’s Llama 3.1-405B, and DeepSeek-V3. And what better way to flex than dropping it on Lunar New Year? But let’s be real  —the real reason? DeepSeek’s dirt-cheap AI models have competitors panicking, leading to price cuts of up to 97% across China’s AI market.

🦢 Jack Dorsey’s fintech company, Block, just unleashed Goose, an open-source AI framework that lets developers build AI assistants without Big Tech breathing down their necks. It plays nice with OpenAI, DeepSeek, and Anthropic, while keeping data private and fully under user control. Block’s already using it to handle boring-but-important tasks like code migrations and dependency management — because why make humans suffer when a bot can do it?

Let's Chat About AI

This is all for today!

Replay to this email with your thoughts.

AI is more than just a buzzword. It’s a shift in how we live and work. And understanding it a bit better means you can make smarter choices about the tech you use every day.

_FEEDBACK