Good morning!

Here's your AI Trends dose, the newsletter that slips through the digital underground like a savvy hacker, uncovering the latest exploits and skirmishes in the world of AI, tech, and innovation. 

Today's clandestine download exposes AI-powered malware 💻 where hackers have gotten chatty with large language models, pulls back the curtain on the data hunger games as tech titans ferociously feed their 🍴 insatiable AI appetites, and reveals Congress's demand that AI pries open the copyright vault. 🔓

Prepare to go off the grid as we infiltrate today's hot topics!🚀

AI-Powered Malware: Hackers Get Chatty with LLMs 🤖💻

Well, well, well...look who decided to enlist some AI assistance for their dastardly schemes. Security researchers have spotted signs that a hacking group is leveraging large language models like ChatGPT to enhance their malware attacks. 

The cybercriminal gang dubbed "TA547" has been firing off phishing emails packed with the Rhadamanthys malware to businesses in Germany. Nothing too novel there. But the real intrigue lies in their malicious PowerShell script containing comments that seem awfully...articulate for your average black hat hacker.

Even hackers are outsourcing work to AI now - first apps, now malware scripts? 🥴

Apparently, the script featured extremely clear code explanations formatted in a way characteristic of LLMs like ChatGPT or Microsoft's Copilot assistant. Security firm Proofpoint theorizes the crooks either straight-up copied AI-generated code or used an LLM tool to rewrite the script with robotic eloquence.

Of course, using AI doesn't necessarily make the malware more potent - TA547's attack was about as menacing as a trick email from a deposed Nigerian prince. But it does showcase how easily accessible generative AI can streamline typically complex cybercriminal workflows.

While still an emerging threat, AI-augmented malware could become increasingly common as criminal outfits look to leverage the latest tech. Sure, ChatGPT itself may refuse to engage in illegal shenanigans. But as we've seen, nothing's stopping hackers from outsourcing the more tedious coding tasks to the AI.  

The one upside? At least when the robots inevitably turn on us, their malware comments will be impeccably clear thanks to prior LLM training. Every cloud has a silver lining!

Data Hunger Games: Tech Titans' Insatiable AI Appetites 📈🤖🍴

Forget scanning books in secret book depot trucks - today's tech giants are going even further to feed their AI model's insatiable appetite for data. According to a New York Times exposé, the likes of OpenAI, Google, and Meta have been bending if not outright breaking copyright laws in their relentless pursuit of training datasets.

From transcribing millions of YouTube videos to discussing methods to vacuum up copyrighted internet content en masse, no social media post, video transcript or e-book is off-limits for these data prospectors. When OpenAI faced a training data shortage for its vaunted GPT models, some employees allegedly greenlit using YouTube content despite the platform's clear prohibitions.  

AI companies when creators ask about legal use of their work: [']);. parse; .error] 🤖

Of course, these corporations claim their liberal data sourcing falls under the nebulous "fair use" clause for research purposes. But unless every YouTuber, author, and viral tweeter pre-approved having their content ingested to create hyper-intelligent AI assistants, that sounds like a piping hot legal rationalization.

It's not just text being gobbled up either - images, voices, you name it. Anything that can teach an AI model new tricks is fair game, individual privacy and consent be damned. Google updated its terms to potentially tap into personal Google Docs, while Meta weighed purchasing book publishers outright. 

On one hand, you can't fault companies for trying to stay competitive in the white-hot AI arms race by any means necessary. Quality training data is the rocket fuel propelling these revolutionary technologies forward. But doing so at the expense of creators' rights and people's privacy is a bridge too far.

As synthetic AI data emerges and lawsuits loom, big tech may soon learn a harsh lesson - cutting ethical corners to rapidly advance AI is a deal they might regret making with the digital devil.

Congress to AI: Open the Copyright Vault 🔓📜

In the rapidly evolving world of generative AI, a new battle is brewing - and this time, it's the lawmakers taking aim at the tech giants. Rep. Adam Schiff has introduced the Generative AI Copyright Disclosure bill, a legislative shot across the bow demanding transparency around the copyrighted materials ingested to train AI models.

If passed, any company unleashing a generative AI system would be required to submit a detailed report to the Copyright Office disclosing exactly what copyrighted works were used in the training data set. No more keeping those model diets a tightly guarded secret.

"Our models were trained on the complete works of Shakespeare, Adam Sandler's entire filmography, and 10 million YouTube makeup tutorials. You're welcome." - AI Companies, probably

The move comes as creators from authors to artists cry foul over their intellectual property potentially being exploited to create ultra-powerful AI without consent or compensation. Tech firms have handwaved these concerns by claiming any copyrighted snippets qualify as fair use for the purpose of research. But this bill could blow that flimsy argument out of the water.

Predictably, groups like the Writers Guild, music industry, and creators' unions are championing the legislation as a vital step in protecting their members' rights. Though the Motion Picture Association's silence suggests some holdouts believe transparency could do more harm than good for certain industries.

Of course, the AI companies themselves are likely scrambling their legal teams as we speak to counter the bill. Disclosing their entire training diet could expose proprietary secrets and provide ammo for a tsunami of copyright lawsuits. That's enough to give any self-respecting AI a crippling anxiety attack.

In the high-stakes tug-of-war between rapidly advancing AI and steadfast copyright protections, Schiff's bill is the latest disruptive force. Whatever happens, it's clear the generative AI revolution won't be an artistic free-for-all - creators are staking their claim.

AI Bites 🍪

Another llama: Meta's Next AI Llama Unleashed This Month 📆🦙🤖

Intel Drops AI Bombshells: Gaudi 3, Open Systems & Big Wins 💥🤖📈

Climbing the Ranks: Cohere's Command R+ Soars 🧗‍♀️🔝🤖

Fasten Your Seatbelts: GPT-4 Turbo's Power Boost Hits OpenAI 🚀✨🤖

Dress to Impress: eBay's AI Fashion Concierge �družžždress🤖

Music to the Next Level: Check out Udio’s Brand New Bag!!

Watch this, thank me later... 👀

Singularity is not quite there yet…

app_source_hub

View more on Instagram

^{The content of this newsletter is for educational and informational purposes only and should not be construed as financial advice. Conduct your own research or consult a financial professional before making investment decisions.}