The Prompt: Former OpenAI CTO Mira Murati Unveils Her New AI Startup

Today, former OpenAI CTO Mira Murati introduced her latest venture: Thinking Machine Labs, a public benefit corporation with a bold mission—making artificial intelligence more accessible and widely capable. After parting ways with OpenAI last September, Murati has assembled a team of talented engineers and researchers, many of whom have worked at leading AI startups like Character AI, Mistral, and, of course, OpenAI.

What makes Thinking Machine Labs stand out? Transparency. The company has committed to sharing its technical research, publishing blog posts, releasing code, and collaborating with the wider AI community. This suggests they may take an open-source approach—a rare move in today’s competitive AI landscape.

Now, let’s dive into the biggest AI news of the day.

BIG PLAYS

The race to build the smartest AI just got even more intense. On Monday, Elon Musk’s AI company, xAI, introduced its latest model: Grok-3—an advanced AI designed to tackle complex questions in science, math, and beyond.

During a livestream on X, Musk shared his grand vision, saying the company’s goal is to “understand the universe… to figure out what’s going on, where the aliens are, and what the meaning of life is.” Bold? Absolutely. But that’s Musk for you.

So, what’s different about Grok-3? Musk claims it was trained with 10 times more computing power than Grok-2, thanks to xAI’s massive "gigafactory of compute" in Memphis. The model has been trained on vast amounts of public data, including information from X (formerly Twitter) and legal documents.

Alongside Grok-3, xAI also introduced DeepSearch, an AI-powered search engine aiming to rival traditional search methods. But is Grok-3 truly a game-changer? OpenAI cofounder and former Tesla executive Andrej Karpathy tested the model and found it comparable to OpenAI’s best offerings—but not without flaws. Some users have also noted that its coding skills fall short compared to other models.

For now, Grok-3 remains untested by independent experts and is available only to paying subscribers. Whether it lives up to the hype remains to be seen.

ETHICS + LAW

A group of major publishers, including Condé Nast, Vox, and The Atlantic, has filed a lawsuit against Cohere, a Canadian AI company valued at $5.5 billion, accusing it of copyright and trademark violations. (Forbes is also part of the lawsuit.)

The publishers claim that Cohere illegally scraped around 4,000 copyrighted articles from the internet to train its AI language models, known as Command. These models allegedly reproduce sections—or sometimes entire articles—word for word, providing users with information without directing them to the original publishers’ websites.

This isn’t the first time an AI company has been accused of misusing copyrighted content. Last year, AI search engine Perplexity faced backlash for republishing copyrighted material from various sources, including Forbes. In response, Forbes issued a cease-and-desist letter, accusing Perplexity of violating copyright laws.

At the heart of this controversy is a growing tension between AI companies and publishers. While AI promises innovation, it also raises ethical and legal concerns about who owns the content fueling these models—and whether creators are being fairly compensated for their work.

AI DEALS OF THE WEEK

Big things are happening in the world of AI!

Figure AI, a company focused on humanoid robots, is in talks to secure a massive $1.5 billion in venture capital, pushing its valuation to a staggering $39.5 billion, according to Bloomberg. What's even more exciting? The company is reportedly in discussions with Meta to create robots that can handle household chores—imagine a future where AI-powered assistants help with everyday tasks at home!

Meanwhile, Luminance, an AI-driven legal tech firm, just raised $75 million in Series C funding. The company is already helping major clients like AMD and National Grid streamline contract creation, negotiation, and analysis—making complex legal work faster and smarter.

And in the AI chip space, Encharge AI has secured $100 million in Series B funding, led by Tiger Global. CEO Naveen Verma, who founded the company from a Princeton lab, has been working on cutting-edge hardware to make large language models run more efficiently. His team’s breakthrough could mean AI models operating seamlessly on personal devices, making AI more accessible and less dependent on cloud computing.

The AI revolution is moving fast, and these deals prove that the future is being built right now.

DEEP DIVE

Elon Musk’s unexpected bid to take over the nonprofit that controls OpenAI created exactly the kind of stir he wanted. As OpenAI CEO Sam Altman and other global leaders gathered in Paris for an AI summit, Musk’s unsolicited $97.4 billion offer put the spotlight back on him—and on his efforts to block OpenAI’s transformation into a for-profit company.

Altman, clearly frustrated, wasted no time in rejecting the offer. Insiders at OpenAI say there’s almost no chance it will go anywhere. But even if Musk doesn’t get what he wants, he’s thrown a wrench into Altman’s plans. By publicly naming a high price for the nonprofit, Musk has raised the stakes, making it much harder for OpenAI to justify a lower valuation.

This bid is the first concrete figure attached to the nonprofit that oversees OpenAI. For the transition to a for-profit model to happen, that entity would need to be bought out and reduced to a minority stakeholder. Until now, reports had pegged its value at around $40 billion, based on a 25% stake in the company. But Musk’s bold $97.4 billion bid puts Altman in a difficult position. As a board member, he now faces growing pressure to sell at that price—or risk looking like he’s undervaluing his own company just to keep control and avoid share dilution.

“If the board turns Musk down, which they almost certainly will, they’re essentially saying these assets are worth more than $97 billion,” a source close to the company told Forbes. “That means if OpenAI’s for-profit side ever tries to buy them out in the future, the nonprofit will have to demand even more—otherwise, the board could be accused of failing in their duty.”

Musk may not have secured the deal, but he’s certainly made things more complicated for Altman.

Model Behavior

Generative AI is giving scammers a powerful new tool to manipulate emotions and deceive people in the world of online dating, Wired reported. Fraudsters are now using AI chatbots to create hundreds of fake conversations and craft convincing profiles on dating apps, making romance scams more widespread and harder to detect.

AI has already started reshaping the way people interact in the dating world. Last year, we shared the story of a man who used ChatGPT to respond to his Tinder matches and even arrange dates on his behalf. While AI can simplify social interactions, it also raises concerns about authenticity and emotional manipulation in modern relationships.