Happy Tuesday. The main thing happening today is OpenAI being sued by the state of California. There’s no show this morning as we’re travelling to Helsinki for Slush but we’ll be live tomorrow from 9:15 GMT. Keep an eye out on X!

Live from etn Slush, Helsinki, 18-19 Nov 2026

What’s Happening Today

OpenAI has been sued by the state of California after its agents went rogue and hacked another company.

California Attorney General Rob Bonta filed the lawsuit in San Francisco on Monday, accusing OpenAI of developing dangerous AI models without adequate safeguards and misleading the public about the risks. It's the first lawsuit brought by a US state over the cyberattacks, with Bonta seeking a court order forcing OpenAI to improve its safety procedures and pay for the alleged violations.

It follows an incident in July, when hundreds of OpenAI agents coordinated an attack on Hugging Face, an AI platform used by developers. Bonta claims the attack was the result of a culture at OpenAI that puts rapid development ahead of safety, adding that it was just one of 80 known incidents. His office launched an investigation in September before bringing the lawsuit this week.

OpenAI spokesperson Drew Pusateri argued that the lawsuit unfairly singles out the company, with Anthropic, Meta and Google also reporting similar behaviour from their models. OpenAI has since strengthened its safeguards.

Daily Op-Ed — Ooh La La, France Have a Second AI Lab

OVHcloud have entered the AI race. CEO Octave Klaba announced this morning that Europe’s leading cloud service provider are releasing an open-weight LLM family dubbed “Olala.”

Their first model, Olala Mini, will be available on Hugging Face from Thursday. It is a “compact,” efficient, 7 billion parameter model. This may sound small compared to Mistral’s trillion-parameter Le Chonk but size isn’t everything - it’s all about how you use it. Olala Mini can be run on cheap CPUs rather than expensive GPU clusters, so SMEs can actually run it in-house. Plus, 7B parameters is more or less in line with other “mini” models.

What has left people wanting more is the context window: only 32,000 tokens. Other similar models typically have context windows of at least 128k tokens. To be fair to the OVHcloud team, this is only the beginning. Klaba has said that Olala V2 will be a 100B+ parameter model.

What is perhaps most interesting is that the model was trained on three public European supercomputers, Jupiter, MeluXina and Jean Zay. This is an example of how state-level investment in compute could pay off: private-public partnerships where the outcome is (eventually) competitive open-weight models. More competitors in the space is good, more European competitors in the space is even better. Though it has started with a trickle, this could be a watershed moment for European AI sovereignty.

— Monty

etn. Insight

Enhanced Games founder Christian Angermayer came on the show and explained why some founders quit after taking ayahuasca, stating that psychedelics “show you who you truly are and what makes you happy”.

Angermayer thinks we're living in a time where “it's cool to be a founder”, much like it was once cool to work at “McKinsey” or go into “investment banking”. But it's a stressful life, and if you're only doing it “performatively”, he believes you can become “deeply unhappy”.

For those who walk away after a psychedelic trip, he argues they haven't really changed. They've just realised “the lie they were living”.

Watch the full clip below.

Continental Breakfast