AI2Day Daily Brief — 7 Sept 2026

Good morning, it is Monday, the seventh of September, and we have five stories for you this morning, starting with a familiar name re-entering a business he helped create.

Travis Kalanick, the co-founder who built Uber and was pushed out as its chief executive in 2017, has finally revealed what his secretive new company, Atoms, intends to do with the one point seven billion dollars it raised this summer. The answer is robotaxis: fleets of self-driving cars operating without a human driver. Andreessen Horowitz led that funding round, and Uber itself has put in one hundred million dollars and held talks about using Atoms technology on its own platform. Atoms has also acquired Pronto, an autonomous vehicle company that previously worked in mining and was once led by Anthony Levandowski, the former Uber self-driving chief. A major hiring push and further acquisitions are reportedly planned. Kalanick called the fundraise unfinished business. The robotaxi market now has one more well-funded competitor with a complicated history in the space.

Complicated is putting it gently. Next, a legal settlement that was supposed to be straightforward is turning messy fast.

Anthropic reached a one point five billion dollar copyright settlement with authors whose books were used without permission to train its Claude AI models. A judge had already ruled that training on copyrighted material is legal under fair use, but that the pirating of those books in the first place was not. The deal gave authors three thousand dollars per pirated title. If a book is still in print the money splits fifty-fifty with the publisher; if rights have reverted to the author they keep the full amount. The problem is that multiple authors are now reporting publishers claiming payment on books whose rights reverted years ago, in some cases as far back as seventeen years. Literary agents, who hold no publishing rights at all, are also reportedly filing claims. The Authors Guild says this looks more like sloppy record-keeping than deliberate fraud, but authors are still having to fight for money a court already said is theirs.

And the broader legal fight over AI training data keeps widening. Two more newsrooms have now gone to court.

The Seattle Times and Newsday filed copyright-infringement lawsuits against OpenAI and Microsoft in early September. Both outlets say their published journalism was scraped without permission and without payment to train AI models, and that those models sometimes reproduce passages almost word for word from their original reporting. Microsoft is named because its Copilot assistant, built into Windows and Microsoft Three Sixty Five, runs on OpenAI technology. Critically, the publishers are not only asking for money. They want courts to destroy the training datasets and the AI models themselves. These two suits sit inside a much larger wave of litigation: nearly four hundred local newspapers have now sued OpenAI and Microsoft over similar claims. The question of what remedy a court might actually order, if any, remains entirely open.

From legal disputes to a social shift. In South Korea, the AI chip boom is reshaping who people want to date, what they study, and what they watch.

For decades, Korean matchmaking put doctors, lawyers and judges at the top of the pecking order. That is changing. The matchmaking firm Gayeon says it has seen a surge in preferences for employees at Samsung Electronics and SK Hynix, the country's two dominant memory-chip makers. The driver is straightforward: money. Last year SK Hynix distributed ten percent of its operating profit to employees, which Reuters calculated worked out to roughly seven hundred million won per person, just over half a million US dollars, more than fifteen times the average Korean annual salary. The cultural ripple is visible in the data. Nine in ten undeclared-major engineering students at Seoul National University chose electrical and computer engineering in twenty twenty-four, up from two in ten the year before. A YouTube dating show featuring Samsung and SK Hynix employees drew around six hundred and fifty thousand views. Chip work now carries status that was previously reserved for the professions.

And finally, Nvidia is no longer just selling the picks and shovels of the AI era. It is buying stakes in the miners too.

Today's sponsor is Zero2Hero: an AI poker coach sits at your table, and you can ask it mid hand whether the player across from you is bluffing. The free tier is real: 50 hands a day against AI opponents, no card. Not a trial. Lifetime access is $99, in US dollars, until 15 September, at zero, the number two, hero, dot A I.

Nvidia's portfolio of equity investments, meaning shares held in other companies, reached ninety-nine billion dollars as of late July twenty twenty-five. Two years earlier that figure was two point two billion. In twenty twenty-six alone the company committed more than forty billion dollars across the AI industry. The single largest category is frontier AI labs, the firms building the most capable models, which have received nearly fifty billion dollars in total from Nvidia. That includes a thirty billion dollar commitment to OpenAI earlier this year as part of that company's one hundred and ten billion dollar funding round. Beyond AI labs, Nvidia holds a stake in Intel now valued at thirty billion dollars and a SpaceX position worth twenty-one billion as of June twenty twenty-five. It has also announced plans to acquire AI startup Hugging Face for twelve point nine billion dollars. Nvidia says these investments help it understand where AI is heading. They also keep it closely tied to the companies that will keep buying its chips.

For the full breakdown of all five stories and everything else from the past week, join us on AI Today Weekly, our longer Monday show, wherever you get your podcasts. That is your AI briefing for today. Every story is at A-I-2-Day dot live. That is A, I, the number two, D-A-Y, dot live. We are back tomorrow morning. If you got something out of this, a thumbs up and a subscribe genuinely helps.