What the Record Shows

Elon Musk's lawsuit against OpenAI entered its third week in an Oakland federal courtroom on Monday, with Microsoft Chief Executive Satya Nadella taking the stand to testify about OpenAI's brief 2023 ouster of Sam Altman. (Source: WSJ CFO Journal)

The trial's most consequential dynamic, though, is not between Musk and OpenAI. It is between Musk and the public record of everything he has said about AI safety over the previous decade. The plaintiff's own statements — talks, posts, founding documents, congressional remarks — are being replayed under oath and functioning, as one observer put it, as a sworn deposition on the industry's safety culture itself. (Source: News Anyway)

The exhibits are not staying confined to the Musk case. This year has brought a documented shift in the legal discourse around AI safety: state attorneys general, federal judges, and product-liability lawyers have begun treating AI defects with the same legal apparatus that handled tobacco, pharmaceuticals, and automotive defects in earlier decades. (Source: Bloomberg Law)

The shift matters because, for the first time, the safety claims AI labs have made publicly are being entered into the kind of record that does not go away. The Musk trial is not just about who controls OpenAI, but about whether the safety theater of the last decade survives cross-examination.

Mexico now has more than 40 million credit cards in circulation, up from fewer than 30 million in 2022 and roughly 25 million in 2018. Banks credit "AI and hyper-personalization of products" for the expansion, telling reporters the technology allows them to extend credit to demographics they used to refuse outright. (Source: The Mexico Political Economist)

The headline number obscures the structure underneath it. Bloomberg, cited in the same report, found that roughly half of Mexico's credit cards go unused, a figure most credibly explained by venture-backed fintechs counting a single customer's digital and physical cards as separate accounts to inflate the growth metrics that keep their funding rounds open. Mexico's government has been forced to pass a law prohibiting financial institutions from charging fees on cards that the customer never asked for.

The cards that are active charge what Mexican regulators consider lawful and most Americans would consider predatory. Total annual cost on consumer cards routinely passes 150% — compared with the 36% APR that triggers public outcry in the United States. Consumer credit grew 7.2% in the first quarter of 2026 while corporate credit shrank, a pattern that points to households using cards to pay for groceries in an economy that is slowing down.

The AI-deployment story in the global north this week was OpenAI's $4 billion Deployment Company and Anthropic's $1.5 billion joint venture with Blackstone and Goldman Sachs, both announced Monday. (Source: Reuters) Forty million cards is what AI deployment in Mexico looked like by the time those announcements landed.

The deployment companies were not informed.

OpenAI announced on Monday the creation of a new $4 billion entity, the OpenAI Deployment Company, with capital from TPG, Bain Capital, Brookfield, and Advent. The new venture, valued at $10 billion and majority-owned by OpenAI, will acquire London-based Tomoro for roughly 150 deployment engineers and existing enterprise clients including Mattel, Tesco, and Virgin Atlantic. (Source: Reuters)

Anthropic announced a parallel $1.5 billion joint venture with Blackstone and Goldman Sachs within the same minute. The two announcements landing 60 seconds apart confirmed what enterprise customers have been signaling for months: the model labs are now selling outcomes, not API access. (Source: Evolving AI Insights)

The private-equity backers on the OpenAI venture got an unusual concession: preferred stock with a guaranteed 17.5% annual return. Defensive structures of that kind are routine in distressed-asset funds. They are not routine in vehicles attached to the most-valued startup in the world. The structure tells you what the labs' own backers privately believe: that enterprise-AI deployment economics do not yet work, and the people writing the checks want their downside priced in before the work begins. (Source: Reuters)

The context in which the new entities are operating, per PwC's 2026 CEO survey released earlier this month: 56% of CEOs report getting nothing from their AI spend, and only a quarter of enterprises have moved more than 40% of pilots into production. Five and a half billion dollars of fresh capital, with downside floors, is what the labs are spending to fix those two numbers. (Source: Evolving AI Insights)